describealign 1.2.0__py3-none-any.whl → 2.0.1__py3-none-any.whl

describealign.py

@@ -1,11 +1,9 @@
+__version__ = '2.0.1'
+
 # combines videos with matching audio files (e.g. audio descriptions)
 # input: video or folder of videos and an audio file or folder of audio files
 # output: videos in a folder "videos_with_ad", with aligned segments of the audio replaced
 # this script aligns the new audio to the video using the video's old audio
-# first, the video's sound and the audio file are both converted to spectrograms
-# second, the two spectrograms are roughly aligned by finding their longest common subsequence
-# third, the rough alignment is denoised through L1-Minimization
-# fourth, the spectrogram alignments determine where the new audio replaces the old
 
 '''
 Copyright (C) 2023  Julian Brown
@@ -28,26 +26,14 @@ VIDEO_EXTENSIONS = set(['mp4', 'mkv', 'avi', 'mov', 'webm', 'm4v', 'flv', 'vob']
 AUDIO_EXTENSIONS = set(['mp3', 'm4a', 'opus', 'wav', 'aac', 'flac', 'ac3', 'mka'])
 PLOT_ALIGNMENT_TO_FILE = True
 
-TIMESTEP_SIZE_SECONDS = .16
-TIMESTEP_OVERLAP_RATIO = .5
+TIMESTEPS_PER_SECOND = 10  # factors must be subset of (2, 3, 5, 7)
+TIMESTEP_SIZE_SECONDS = 1. / TIMESTEPS_PER_SECOND
 AUDIO_SAMPLE_RATE = 44100
-MEL_COEFFS_PER_TIMESTEP = 25
-DITHER_PERIOD_STEPS = 60
-MIN_CORR_FOR_TOKEN_MATCH = .6
-GAP_START_COST = 1.0
-GAP_EXTEND_COST = -.01
-GAP_EXTEND_DIAG_BONUS = -.01
-SKIP_MATCH_COST = .1
+DITHER_PERIOD_STEPS = 10
 MAX_RATE_RATIO_DIFF_ALIGN = .1
-PREF_CUT_AT_GAPS_FACTOR = 5
 MIN_DURATION_TO_REPLACE_SECONDS = 2
-MIN_START_END_SYNC_TIME_SECONDS = 2
-MAX_START_END_SYNC_ERR_SECONDS = .2
-MAX_RATE_RATIO_DIFF_BOOST = .003
-MIN_DESC_DURATION = .5
-MAX_GAP_IN_DESC_SEC = 1.5
 JUST_NOTICEABLE_DIFF_IN_FREQ_RATIO = .005
-CATCHUP_RATE = 5
+MIN_STRETCH_OFFSET = 30
 
 if PLOT_ALIGNMENT_TO_FILE:
   import matplotlib.pyplot as plt
@@ -64,42 +50,41 @@ import numpy as np
 import ffmpeg
 import platformdirs
 import static_ffmpeg
-import python_speech_features as psf
 import scipy.signal
 import scipy.optimize
 import scipy.interpolate
-import scipy.ndimage as nd
 import scipy.sparse
-import pytsmod
 import configparser
 import traceback
 import multiprocessing
 import platform
+import natsort
+from collections import defaultdict
+from sortedcontainers import SortedList
+import hashlib
+
+try:
+  import wx
+  gui_font = (11, wx.FONTFAMILY_SWISS, wx.FONTSTYLE_NORMAL, wx.FONTWEIGHT_NORMAL, False, "Arial")
+except ImportError:
+  wx = None
+
+gui_update_interval_ms = 100
+gui_background_color_dark = (28, 30, 35)
+gui_background_color_light = (170, 182, 211)
 
 IS_RUNNING_WINDOWS = platform.system() == 'Windows'
 if IS_RUNNING_WINDOWS:
-  import PySimpleGUIWx as sg
   default_output_dir = 'videos_with_ad'
   default_alignment_dir = 'alignment_plots'
 else:
-  import PySimpleGUIQt as sg
   default_output_dir = os.path.expanduser('~') + '/videos_with_ad'
   default_alignment_dir = os.path.expanduser('~') + '/alignment_plots'
 
-def display(text, func=None):
-  if func:
-    func(text)
-  print(text)
-
-def throw_runtime_error(text, func=None):
-  if func:
-    func(text)
-  raise RuntimeError(text)
-
-def ensure_folders_exist(dirs, display_func=None):
+def ensure_folders_exist(dirs):
   for dir in dirs:
     if not os.path.isdir(dir):
-      display(f"Directory not found, creating it: {dir}", display_func)
+      print(f"Directory not found, creating it: {dir}")
       os.makedirs(dir)
 
 def get_sorted_filenames(path, extensions, alt_extensions=set([])):
@@ -127,347 +112,59 @@ def get_sorted_filenames(path, extensions, alt_extensions=set([])):
                  "Or maybe you need to add a new extension to this script's regex?",
                  f"valid extensions for this input are:\n  {extensions}"]
     raise RuntimeError("\n".join(error_msg))
-  files = sorted(files)
-  file_types = [0 if os.path.splitext(file)[1][1:] in extensions else 1 for file in files]
-  return files, file_types
-
-# read audio from file with ffmpeg and convert to numpy array
-def parse_audio_from_file(media_file):
-  media_stream, _ = (ffmpeg
-    .input(media_file)
-    .output('-', format='s16le', acodec='pcm_s16le', ac=2, ar=AUDIO_SAMPLE_RATE, loglevel='fatal')
-    .run(capture_stdout=True, cmd=get_ffmpeg())
-  )
-  media_arr = np.frombuffer(media_stream, np.int16).astype(np.float32).reshape((-1,2)).T
-  return media_arr
-
-# tokenize audio by transforming with a mel-frequency cepstrum (MFC)
-def tokenize_audio(media_arr, rate=1):
-  step_size_samples = psf.sigproc.round_half_up(TIMESTEP_SIZE_SECONDS * rate * AUDIO_SAMPLE_RATE)
-  window_size_seconds = TIMESTEP_SIZE_SECONDS / TIMESTEP_OVERLAP_RATIO
-  window_size_samples = psf.sigproc.round_half_up(window_size_seconds * AUDIO_SAMPLE_RATE)
-  fft_size_samples = 2**int(np.ceil(np.log2(window_size_samples)))
-  get_mfcc = lambda arr: psf.mfcc(np.mean(arr, axis=0),
-                                  samplerate=AUDIO_SAMPLE_RATE,
-                                  winlen=window_size_seconds,
-                                  winstep=TIMESTEP_SIZE_SECONDS * rate,
-                                  numcep=MEL_COEFFS_PER_TIMESTEP,
-                                  nfilt=MEL_COEFFS_PER_TIMESTEP * 2,
-                                  nfft=fft_size_samples,
-                                  winfunc=scipy.signal.windows.hann)
-  num_timesteps = max(1, ((media_arr.shape[1] - window_size_samples - 1) // step_size_samples) + 2)
-  media_spec = np.zeros((num_timesteps, MEL_COEFFS_PER_TIMESTEP))
-  chunk_size = 1000
-  for chunk_index in np.arange(0, num_timesteps, chunk_size):
-    chunk_bounds_samples = ((chunk_index                 ) * step_size_samples,
-                            (chunk_index + chunk_size - 1) * step_size_samples + window_size_samples)
-    media_spec[chunk_index:chunk_index+chunk_size] = get_mfcc(media_arr[:,slice(*chunk_bounds_samples)])
-  '''
-  # alternate python library's MFC implementation
-  import librosa
-  media_spec = librosa.feature.mfcc(y=np.mean(media_arr, axis=0),
-                                    sr=AUDIO_SAMPLE_RATE,
-                                    n_mfcc=MEL_COEFFS_PER_TIMESTEP,
-                                    lifter=22,
-                                    n_fft=fft_size_samples,
-                                    hop_length=step_size_samples,
-                                    win_length=window_size_samples,
-                                    window=scipy.signal.windows.hann).T
-  num_timesteps = media_spec.shape[0]
-  '''
-  timings_samples = window_size_samples/2. + step_size_samples * np.arange(num_timesteps)
-  timings_seconds = timings_samples / AUDIO_SAMPLE_RATE
-  return media_spec, timings_seconds
-
-# same as tokenize_audio, but dithering the MFC window timings
-# this allows for finer alignment by ameliorating discretization error
-def tokenize_audio_dither(media_arr, slow_timings):
-  # choose a relative step size slightly less than 1 to ameliorate quantization error
-  # maximize alignment accuracy by using least approximable number with desired period
-  # this is the continued fraction [0;1,N-2,1,1,1,...], where the trailing ones give phi
-  fast_rate = 1. / (1 + 1. / (DITHER_PERIOD_STEPS - 2 + (np.sqrt(5) + 1) / 2.))
-  fast_spec, fast_timings = tokenize_audio(media_arr, fast_rate)
-  
-  # prevent drift in difficult to align segments (e.g. describer speaking or quiet/droning segments)
-  # by approximately equalizing the number of tokens per unit time between dithered and undithered
-  # the dithered audio will have ~(1 + 1 / DITHER_PERIOD_STEPS) times as many tokens, so
-  # this can be accomplished by simply deleting a token every DITHER_PERIOD_STEPS tokens
-  fast_spec = np.delete(fast_spec, slice(DITHER_PERIOD_STEPS // 2, None, DITHER_PERIOD_STEPS), axis=0)
-  fast_timings = np.delete(fast_timings, slice(DITHER_PERIOD_STEPS // 2, None, DITHER_PERIOD_STEPS))
-  return fast_spec, fast_timings
-
-# normalize along both time and frequency axes to allow comparing tokens by correlation
-def normalize_spec(media_spec_raw, axes=(0,1)):
-  media_spec = media_spec_raw.copy()
-  for axis in axes:
-    norm_func = np.std if axis == 0 else np.linalg.norm
-    media_spec = media_spec - np.mean(media_spec, axis=axis, keepdims=True)
-    media_spec = media_spec/(norm_func(media_spec,axis=axis,keepdims=True)+1e-10)
-  return media_spec
-
-# vectorized implementation of the Wagner–Fischer (Longest Common Subsequence) algorithm
-# modified to include affine gap penalties and skip+match options (i.e. knight's moves)
-# gaps are necessary when parts are cut out of the audio description (e.g. cut credits)
-# or when the audio description includes a commercial break or an extra scene
-# the skip+match option allows for micro-adjustments without eating the full gap penalty
-# skip+match is primarily useful in maintaining alignment when the rates differ slightly
-def rough_align(video_spec, audio_desc_spec, video_timings, audio_desc_timings):
-  pred_map = {0:lambda node: (0, node[1]-1, node[2]-1),
-              1:lambda node: (0, node[1]-2, node[2]-1),
-              2:lambda node: (0, node[1]-1, node[2]-2),
-              3:lambda node: (1, node[1]-1, node[2]-1),
-              4:lambda node: (0, node[1]  , node[2]  ),
-              5:lambda node: (1, node[1]-1, node[2]  ),
-              6:lambda node: (1, node[1]-1, node[2]-1),
-              7:lambda node: (1, node[1]  , node[2]-1)}
-  pred_matrix = np.zeros((2, audio_desc_spec.shape[0], video_spec.shape[0]), dtype=np.uint8)
-  pred_matrix[0,1:,:2] = 0
-  pred_matrix[1,1:,:2] = 4
-  pred_matrix[:,0,:2] = [0,5]
-  path_corrs_match = np.zeros((3, video_spec.shape[0]))
-  path_corrs_gap = np.zeros((3, video_spec.shape[0]))
-  corrs = np.zeros((3, video_spec.shape[0]))
-  corrs[:,:] = np.roll(np.dot(video_spec, audio_desc_spec[0]), 1)[None,:]
-  for i in range(audio_desc_spec.shape[0]):
-    i_mod = i % 3
-    match_pred_corrs = np.hstack([path_corrs_match[i_mod-1][1:-1][:,None],
-                                  path_corrs_match[i_mod-2][1:-1][:,None] - SKIP_MATCH_COST,
-                                  path_corrs_match[i_mod-1][0:-2][:,None] - SKIP_MATCH_COST,
-                                  path_corrs_gap[  i_mod-1][1:-1][:,None]])
-    pred_matrix[0][i][2:] = np.argmax(match_pred_corrs, axis=1)
-    path_corrs_match[i_mod][2:] = np.take_along_axis(match_pred_corrs, pred_matrix[0][i][2:,None], axis=1).T
-    corrs = np.roll(corrs, -1, axis=1)
-    corrs[(i_mod+1)%3,:] = np.roll(np.dot(video_spec, audio_desc_spec[min(audio_desc_spec.shape[0]-1,i+1)]), 1)
-    fisher_infos = (2 * corrs[i_mod] - corrs[i_mod-1] - corrs[(i_mod+1)%3]) / min(.2, TIMESTEP_SIZE_SECONDS)
-    fisher_infos[fisher_infos < 0] = 0
-    fisher_infos[fisher_infos > 10] = 10
-    row_corrs = np.maximum(0, corrs[i_mod][2:] - MIN_CORR_FOR_TOKEN_MATCH)
-    path_corrs_match[i_mod][2:] += row_corrs * (fisher_infos[2:] / 5)
-    gap_pred_corrs = np.hstack([path_corrs_match[i_mod][2:  ][:,None] - GAP_START_COST,
-                                path_corrs_gap[i_mod-1][2:  ][:,None],
-                                path_corrs_gap[i_mod-1][1:-1][:,None] - GAP_EXTEND_DIAG_BONUS - \
-                                                                        GAP_EXTEND_COST])
-    pred_matrix[1][i][2:] = np.argmax(gap_pred_corrs, axis=1)
-    path_corrs_gap_no_col_skip = np.take_along_axis(gap_pred_corrs, pred_matrix[1][i][2:,None], axis=1).flat
-    pred_matrix[1][i][2:] += 4
-    path_corrs_gap[i_mod][2:] = np.maximum.accumulate(path_corrs_gap_no_col_skip + \
-                                                      GAP_EXTEND_COST * np.arange(video_spec.shape[0]-2)) - \
-                                                      GAP_EXTEND_COST * np.arange(video_spec.shape[0]-2)
-    pred_matrix[1][i][2:][path_corrs_gap[i_mod][2:] > path_corrs_gap_no_col_skip] = 7
-    path_corrs_gap[i_mod][2:] -= GAP_EXTEND_COST
-  
-  # reconstruct optimal path by following predecessors backwards through the table
-  end_node_layer = np.argmax([path_corrs_match[i_mod,-1],
-                              path_corrs_gap[  i_mod,-1]])
-  cur_node = (end_node_layer, audio_desc_spec.shape[0]-1, video_spec.shape[0]-1)
-  get_predecessor = lambda node: pred_map[pred_matrix[node]](node)
-  path = []
-  visited = set()
-  while min(cur_node[1:]) >= 0:
-    cur_node, last_node = get_predecessor(cur_node), cur_node
-    # failsafe to prevent an infinite loop that should never happen anyways
-    if cur_node in visited:
-      break
-    visited.add(cur_node)
-    if last_node[0] == 0:
-      path.append(last_node[1:])
-  path = path[::-1]
-  
-  # determine how much information this node gives about the alignment
-  # a larger double derivative means more precise timing information
-  # sudden noises give more timing information than droning sounds
-  def get_fisher_info(node):
-    i,j = node
-    if node[0] >= audio_desc_spec.shape[0]-1 or \
-       node[1] >= video_spec.shape[0]-1 or \
-       min(node) <= 0:
-      return 0
-    info = 2*np.dot(audio_desc_spec[i  ],video_spec[j  ]) - \
-             np.dot(audio_desc_spec[i-1],video_spec[j+1]) - \
-             np.dot(audio_desc_spec[i+1],video_spec[j-1])
-    info /= min(.2, TIMESTEP_SIZE_SECONDS)
-    return info
-  
-  # the quality of a node combines the correlation of its tokens
-  # with how precisely the match is localized in time
-  def get_match_quality(node):
-    # correlations are between -1 and 1, as all tokens have unit norm
-    token_correlation = np.dot(audio_desc_spec[node[0]],video_spec[node[1]])
-    fisher_info = min(max(0, get_fisher_info(node)), 10)
-    return max(0, token_correlation - MIN_CORR_FOR_TOKEN_MATCH) * (fisher_info / 5)
-  
-  # filter out low match quality nodes from LCS path
-  quals = [get_match_quality(node) for node in path]
-  if len(quals) == 0 or max(quals) <= 0:
-    raise RuntimeError("Rough alignment failed, are the input files mismatched?")
-  path, quals = zip(*[(path, qual) for (path, qual) in zip(path, quals) if qual > 0])
-  
-  # convert units of path nodes from timesteps to seconds
-  path = [(audio_desc_timings[i], video_timings[j]) for (i,j) in path]
-  
-  return path, quals
+  files = natsort.os_sorted(files)
+  has_alt_extensions = [0 if os.path.splitext(file)[1][1:] in extensions else 1 for file in files]
+  return files, has_alt_extensions
 
-# chunk path segments of similar slope into clips
-# a clip has the form: (start_index, end_index)
-def chunk_path(smooth_path, tol):
-  x,y = zip(*smooth_path)
-  slopes = np.diff(y) / np.diff(x)
-  median_slope = np.median(slopes)
-  slope_changes = np.diff(slopes)
-  breaks = np.where(np.abs(slope_changes) > tol)[0] + 1
-  breaks = [0] + list(breaks) + [len(x)-1]
-  clips = list(zip(breaks[:-1], breaks[1:]))
-  return clips, median_slope, slopes
+# ffmpeg command error handler
+def run_ffmpeg_command(command, err_msg):
+  try:
+    return command.run(capture_stdout=True, capture_stderr=True, cmd=get_ffmpeg())
+  except ffmpeg.Error as e:
+    print("  ERROR: ffmpeg failed to " + err_msg)
+    print("FFmpeg error:")
+    print(e.stderr.decode('utf-8'))
+    raise
 
-# find piece-wise linear alignment that minimizes the weighted combination of
-# total absolute error at each node and total absolute slope change of the fit
-# distance between nodes and the fit (i.e. errors) are weighted by node quality
-# absolute slope changes are differences between the slopes of adjacent fit lines
-# slope changes are weighted much more than node errors to smooth out noise
-# the main source of noise is rough alignment drift while the describer is speaking
-def smooth_align(path, quals, smoothness):
-  # rotate basis to make vertical and horizontal slopes "cost" the same
-  # the new horizontal axis is x+y and the new vertical is -x+y
-  # Wagner–Fischer gives monotonically increasing nodes, so 0 <= slope < inf
-  # after this transformation, we instead have -1 <= slope < 1
-  # perfectly matching audio has pre-transformation slope = 1
-  # after this transformation, it instead has slope = 0
-  rotated_path = [(x+y,-x+y) for x,y in path]
-  
-  # stretch the x axis to make all slopes "cost" nearly the same
-  # without this, small changes to the slope at slope = +/-1
-  # cost sqrt(2) times as much as small changes at slope = 0
-  # by stretching, we limit the range of slopes to within +/- 1/x_stretch_factor
-  # the small angle approximation means these slopes all cost roughly the same
-  x_stretch_factor = 10.
-  rotated_stretched_path = [(x_stretch_factor*x,y) for x,y in rotated_path]
-  
-  # L1-Minimization to solve the alignment problem using a linear program
-  # the absolute value functions needed for "absolute error" can be represented
-  # in a linear program by splitting variables into positive and negative pieces
-  # and constraining each to be positive (done by default in scipy's linprog)
-  # x is fit_err_pos, fit_err_neg, slope_change_pos, slope_change_neg
-  # fit_err[i] = path[i][1] - y_fit[i]
-  # slope_change[i] = (y_fit[i+2] - y_fit[i+1])/(path[i+2][0] - path[i+1][0]) - \
-  #                   (y_fit[i+1] - y_fit[i  ])/(path[i+1][0] - path[i  ][0])
-  # this can be rewritten in terms of fit_err by re-arranging the 1st equation:
-  #   y_fit[i] = path[i][1] - fit_err[i]
-  # this gives:
-  #   slope_change[i] = path_half[i] - fit_err_half[i]
-  #   where each half is just the original equation but y_fit is swapped out
-  # the slope_change variables can then be set using equality constraints
-  num_fit_points = len(rotated_stretched_path)
-  x,y = [np.array(arr) for arr in zip(*rotated_stretched_path)]
-  x_diffs = np.diff(x, prepend=[-10**10], append=[10**10])
-  y_diffs = np.diff(y, prepend=[  0    ], append=[ 0    ])
-  slope_change_magnitudes = np.abs(np.diff(y_diffs/x_diffs)) * x_stretch_factor
-  slope_change_locations = (slope_change_magnitudes > MAX_RATE_RATIO_DIFF_ALIGN)
-  slope_change_locations[1:-1] *= (np.abs(y[2:] - y[:-2]) > 5)
-  slope_change_costs = np.full(num_fit_points, smoothness / float(TIMESTEP_SIZE_SECONDS))
-  slope_change_costs[slope_change_locations] /= PREF_CUT_AT_GAPS_FACTOR
-  c = np.hstack([quals,
-                 quals,
-                 slope_change_costs * x_stretch_factor,
-                 slope_change_costs * x_stretch_factor])
-  fit_err_coeffs = scipy.sparse.diags([ 1. / x_diffs[:-1],
-                                       -1. / x_diffs[:-1] - 1. / x_diffs[1:],
-                                                            1. / x_diffs[1:]],
-                                      offsets=[0,1,2],
-                                      shape=(num_fit_points, num_fit_points + 2)).tocsc()[:,1:-1]
-  A_eq = scipy.sparse.hstack([ fit_err_coeffs,
-                              -fit_err_coeffs,
-                               scipy.sparse.eye(num_fit_points),
-                              -scipy.sparse.eye(num_fit_points)])
-  b_eq = y_diffs[1:  ] / x_diffs[1:  ] - \
-         y_diffs[ :-1] / x_diffs[ :-1]
-  fit = scipy.optimize.linprog(c, A_eq=A_eq, b_eq=b_eq, method='highs-ds')
-  # if dual simplex solver encounters numerical problems, retry with interior point solver
-  if not fit.success and fit.status == 4:
-    fit = scipy.optimize.linprog(c, A_eq=A_eq, b_eq=b_eq, method='highs-ipm')
-  if not fit.success:
-    print(fit)
-    raise RuntimeError("Smooth Alignment L1-Min Optimization Failed!")
-  
-  # combine fit_err_pos and fit_err_neg
-  fit_err = fit.x[:num_fit_points] - fit.x[num_fit_points:2*num_fit_points]
-  
-  # subtract fit errors from nodes to retrieve the smooth fit's coordinates
-  # also, unstretch x axis and rotate basis back, reversing the affine pre-processing
-  smooth_path = [(((x / x_stretch_factor) - y) / 2.,
-                  ((x / x_stretch_factor) + y) / 2.) for x,y in zip(x, y - fit_err)]
-  
-  # clip off start/end of replacement audio if it doesn't match or isn't aligned
-  # without this, describer intro/outro skips can cause mismatches at the start/end
-  # the problem would be localized and just means audio might not match video at the start/end
-  # instead we just keep the original video's audio in those segments if mismatches are detected
-  # if instead the first few or last few nodes are well-aligned, that edge is marked as synced
-  # during audio replacement, synced edges will be extended backwards/forwards as far as possible
-  # this is useful when the describer begins talking immediately (or before any alignable audio)
-  # or when the describer continues speaking until the end (or no more alignable audio remains)
-  # otherwise, the mismatch would result in the describer's voice not replacing audio in that part
-  max_sync_err = MAX_START_END_SYNC_ERR_SECONDS
-  smoothing_std = MIN_START_END_SYNC_TIME_SECONDS / (2. * TIMESTEP_SIZE_SECONDS)
-  smoothed_fit_err = nd.gaussian_filter(np.abs(fit_err), sigma=smoothing_std)
-  smooth_err_path = zip(smoothed_fit_err, smooth_path)
-  old_length = num_fit_points
-  smooth_err_path = list(itertools.dropwhile(lambda x: x[0] > max_sync_err, smooth_err_path))[::-1]
-  is_synced_at_start = len(smooth_err_path) == old_length
-  old_length = len(smooth_err_path)
-  smooth_err_path = list(itertools.dropwhile(lambda x: x[0] > max_sync_err, smooth_err_path))[::-1]
-  is_synced_at_end = len(smooth_err_path) == old_length
-  _, smooth_path = zip(*smooth_err_path)
-  smooth_path = list(smooth_path)
-  if is_synced_at_start:
-    slope = (smooth_path[1][1] - smooth_path[0][1]) / (smooth_path[1][0] - smooth_path[0][0])
-    smooth_path.insert(0, (-10e10, -10e10 * slope))
-  if is_synced_at_end:
-    slope = (smooth_path[-1][1] - smooth_path[-2][1]) / (smooth_path[-1][0] - smooth_path[-2][0])
-    smooth_path.append((10e10, 10e10 * slope))
-  
-  clips, median_slope, slopes = chunk_path(smooth_path, tol=1e-7)
-  
-  # assemble clips with slopes within the rate tolerance into runs
-  runs, run = [], []
-  bad_clips = []
-  for clip in clips:
-    if np.abs(median_slope-slopes[clip[0]]) > MAX_RATE_RATIO_DIFF_ALIGN:
-      if len(run) > 0:
-        runs.append(run)
-        run = []
-      bad_clips.append(clip)
-      continue
-    run.append(clip)
-  if len(run) > 0:
-    runs.append(run)
-  
-  return smooth_path, runs, bad_clips, clips
+def run_async_ffmpeg_command(command, media_arr, err_msg):
+  try:
+    ffmpeg_caller = command.run_async(pipe_stdin=True, quiet=True, cmd=get_ffmpeg())
+    out, err = ffmpeg_caller.communicate(media_arr.astype(np.int16).T.tobytes())
+    if len(err) > 0:
+      print("  ERROR: ffmpeg failed to " + err_msg)
+      print("FFmpeg error:")
+      print(err.decode('utf-8'))
+      raise ChildProcessError('FFmpeg error.')
+  except ffmpeg.Error as e:
+    print("  ERROR: ffmpeg failed to " + err_msg)
+    print("FFmpeg error:")
+    print(e.stderr.decode('utf-8'))
+    raise
 
-# if the start or end were marked as synced during smooth alignment then
-# extend that alignment to the edge (i.e. to the start/end of the audio)
-def cap_synced_end_points(smooth_path, video_arr, audio_desc_arr):
-  if smooth_path[0][0] < -10e9:
-    slope = smooth_path[0][1] / smooth_path[0][0]
-    new_start_point = (0, smooth_path[1][1] - smooth_path[1][0] * slope)
-    if new_start_point[1] < 0:
-      new_start_point = (smooth_path[1][0] - smooth_path[1][1] / slope, 0)
-    smooth_path[0] = new_start_point
-  if smooth_path[-1][0] > 10e9:
-    video_runtime = (video_arr.shape[1] - 2.) / AUDIO_SAMPLE_RATE
-    audio_runtime = (audio_desc_arr.shape[1] - 2.) / AUDIO_SAMPLE_RATE
-    slope = smooth_path[-1][1] / smooth_path[-1][0]
-    new_end_point = (audio_runtime, smooth_path[-2][1] + (audio_runtime - smooth_path[-2][0]) * slope)
-    if new_end_point[1] > video_runtime:
-      new_end_point = (smooth_path[-2][0] + (video_runtime - smooth_path[-2][1]) / slope, video_runtime)
-    smooth_path[-1] = new_end_point
+# read audio from file with ffmpeg and convert to numpy array
+def parse_audio_from_file(media_file, num_channels=2):
+  # retrieve only the first audio track, injecting silence/trimming to force timestamps to match up
+  # for example, when the video starts before the audio this fills that starting gap with silence
+  ffmpeg_command = ffmpeg.input(media_file).output('-', format='s16le', acodec='pcm_s16le',
+                                                   af='aresample=async=1:first_pts=0', map='0:a:0',
+                                                   ac=num_channels, ar=AUDIO_SAMPLE_RATE, loglevel='error')
+  media_stream, _ = run_ffmpeg_command(ffmpeg_command, f"parse audio from input file: {media_file}")
+  # media_arr = np.frombuffer(media_stream, np.int16).astype(np.float32).reshape((-1, num_channels)).T
+  media_arr = np.frombuffer(media_stream, np.int16).astype(np.float16).reshape((-1, num_channels)).T
+  return media_arr
 
-# visualize both the rough and smooth alignments
-def plot_alignment(plot_filename_no_ext, path, smooth_path, quals, runs,
-                   bad_clips, ad_timings, similarity_percent):
+def plot_alignment(plot_filename_no_ext, path, audio_times, video_times, similarity_percent,
+                   median_slope, stretch_audio, no_pitch_correction):
+  downsample = 20
+  path = path[::downsample]
+  video_times_full, audio_times_full, cluster_indices, quals, cum_quals = path.T
   scatter_color = [.2,.4,.8]
   lcs_rgba = np.zeros((len(quals),4))
   lcs_rgba[:,:3] = np.array(scatter_color)[None,:]
-  lcs_rgba[:,3] = np.minimum(1, np.array(quals) * 500. / len(quals))
-  audio_times, video_times = np.array(path).T.reshape((2,-1))
+  lcs_rgba[:,3] = np.clip(quals * 400. / len(quals), 0, 1)
+  audio_offsets = audio_times_full - video_times_full
+  plt.switch_backend('Agg')
+  plt.scatter(video_times_full / 60., audio_offsets, s=3, c=lcs_rgba, label='Matches')
   audio_offsets = audio_times - video_times
   def expand_limits(start, end, ratio=.01):
     average = (end + start) / 2.
@@ -475,63 +172,59 @@ def plot_alignment(plot_filename_no_ext, path, smooth_path, quals, runs,
     half_diff *= (1 + ratio)
     return (average - half_diff, average + half_diff)
   plt.xlim(expand_limits(*(0, np.max(video_times) / 60.)))
-  plt.ylim(expand_limits(*(np.min(audio_offsets) - TIMESTEP_SIZE_SECONDS / 2.,
-                          np.max(audio_offsets) + TIMESTEP_SIZE_SECONDS / 2.)))
-  plt.scatter(video_times / 60., audio_offsets, s=3, c=lcs_rgba, label='LCS Matches')
-  audio_times, video_times = np.array(smooth_path).T.reshape((2,-1))
-  audio_offsets = audio_times - video_times
-  if ad_timings is None:
+  plt.ylim(expand_limits(*(np.min(audio_offsets) - 10 * TIMESTEP_SIZE_SECONDS,
+                           np.max(audio_offsets) + 10 * TIMESTEP_SIZE_SECONDS), .05))
+  if stretch_audio:
     plt.plot(video_times / 60., audio_offsets, 'r-', lw=.5, label='Replaced Audio')
-    bad_path = []
-    for clip in bad_clips:
-      bad_path.extend(smooth_path[clip[0]:clip[1]+1])
-      bad_path.append((smooth_path[clip[1]][0] + 1e-10, np.nan))
-    audio_times, video_times = np.array(bad_path).T.reshape((2,-1))
-    audio_offsets = audio_times - video_times
-    if len(audio_offsets) > 0:
-      plt.plot(video_times / 60., audio_offsets, 'c-', lw=1, label='Original Audio')
+    audio_times_unreplaced = []
+    video_times_unreplaced = []
+    for i in range(len(video_times) - 1):
+      slope = (audio_times[i+1] - audio_times[i]) / (video_times[i+1] - video_times[i])
+      if abs(1 - slope) > MAX_RATE_RATIO_DIFF_ALIGN:
+        video_times_unreplaced.extend(video_times[i:i+2])
+        audio_times_unreplaced.extend(audio_times[i:i+2])
+        video_times_unreplaced.append(video_times[i+1])
+        audio_times_unreplaced.append(np.nan)
+    if len(video_times_unreplaced) > 0:
+      video_times_unreplaced = np.array(video_times_unreplaced)
+      audio_times_unreplaced = np.array(audio_times_unreplaced)
+      audio_offsets = audio_times_unreplaced - video_times_unreplaced
+      plt.plot(video_times_unreplaced / 60., audio_offsets, 'c-', lw=1, label='Original Audio')
   else:
-    interp = scipy.interpolate.interp1d(video_times, audio_offsets,
-                                        fill_value = np.inf,
-                                        bounds_error = False, assume_sorted = True)
-    plt.plot(video_times / 60., audio_offsets, 'c-', lw=.5, label='Original Audio')
-    video_times = ad_timings
-    audio_offsets = interp(ad_timings)
-    if len(audio_offsets) > 0:
-      plt.plot(video_times / 60., audio_offsets, 'r-', lw=1, label='Replaced Audio')
-  plt.xlabel('Video Time (minutes)')
-  plt.ylabel('Audio Description Offset (seconds)')
+    plt.plot(video_times / 60., audio_offsets, 'r-', lw=1, label='Combined Media')
+  plt.xlabel('Original Video Time (minutes)')
+  plt.ylabel('Original Audio Description Offset (seconds behind video)')
   plt.title(f"Alignment - Media Similarity {similarity_percent:.2f}%")
   plt.legend().legend_handles[0].set_color(scatter_color)
   plt.tight_layout()
   plt.savefig(plot_filename_no_ext + '.png', dpi=400)
   plt.clf()
-  
   with open(plot_filename_no_ext + '.txt', 'w') as file:
-    rough_clips, median_slope, _ = chunk_path(smooth_path, tol=2e-2)
-    video_offset = np.diff(smooth_path[rough_clips[0][0]])[0]
+    parameters = {'stretch_audio':stretch_audio, 'no_pitch_correction':no_pitch_correction}
+    print(f"Parameters: {parameters}", file=file)
+    this_script_path = os.path.abspath(__file__)
+    print(f"Version Hash: {get_version_hash(this_script_path)}", file=file)
+    video_offset = video_times[0] - audio_times[0]
     print(f"Input file similarity: {similarity_percent:.2f}%", file=file)
     print("Main changes needed to video to align it to audio input:", file=file)
     print(f"Start Offset: {-video_offset:.2f} seconds", file=file)
     print(f"Median Rate Change: {(median_slope-1.)*100:.2f}%", file=file)
-    for clip_start, clip_end in rough_clips:
-      audio_desc_start, video_start = smooth_path[clip_start]
-      audio_desc_end, video_end = smooth_path[clip_end]
-      slope = (video_end - video_start) / (audio_desc_end - audio_desc_start)
+    for i in range(len(video_times) - 1):
+      slope = (video_times[i+1] - video_times[i]) / (audio_times[i+1] - audio_times[i])
       def str_from_time(seconds):
         minutes, seconds = divmod(seconds, 60)
         hours, minutes = divmod(minutes, 60)
-        return f"{hours:2.0f}:{minutes:02.0f}:{seconds:05.2f}"
-      print(f"Rate change of {(slope-1.)*100:6.1f}% from {str_from_time(video_start)} to " + \
-            f"{str_from_time(video_end)} aligning with audio from " + \
-            f"{str_from_time(audio_desc_start)} to {str_from_time(audio_desc_end)}", file=file)
+        return f"{hours:2.0f}:{minutes:02.0f}:{seconds:06.3f}"
+      print(f"Rate change of {(slope-1.)*100:8.1f}% from {str_from_time(video_times[i])} to " + \
+            f"{str_from_time(video_times[i+1])} aligning with audio from " + \
+            f"{str_from_time(audio_times[i])} to {str_from_time(audio_times[i+1])}", file=file)
 
 # use the smooth alignment to replace runs of video sound with corresponding described audio
-def replace_aligned_segments(video_arr, audio_desc_arr, smooth_path, runs, no_pitch_correction=False):
+def replace_aligned_segments(video_arr, audio_desc_arr, audio_desc_times, video_times, no_pitch_correction):
   # perform quadratic interpolation of the audio description's waveform
   # this allows it to be stretched to match the corresponding video segment
   def audio_desc_arr_interp(samples):
-    chunk_size = 10**7
+    chunk_size = 10**5
     interpolated_chunks = []
     for chunk in (samples[i:i+chunk_size] for i in range(0, len(samples), chunk_size)):
       interp_bounds = (max(int(chunk[0]-2), 0),
@@ -540,215 +233,196 @@ def replace_aligned_segments(video_arr, audio_desc_arr, smooth_path, runs, no_pi
                                           audio_desc_arr[:,slice(*interp_bounds)],
                                           copy=False, bounds_error=False, fill_value=0,
                                           kind='quadratic', assume_sorted=True)
-      interpolated_chunks.append(interp(chunk).astype(np.float32))
+      interpolated_chunks.append(interp(chunk).astype(np.float16))
     return np.hstack(interpolated_chunks)
   
-  # construct a stretched audio description waveform using the quadratic interpolator
-  def get_interped_segment(run, interp):
-    segment = []
-    for clip in run:
-      num_samples = int(y[clip[1]] * AUDIO_SAMPLE_RATE) - \
-                    int(y[clip[0]] * AUDIO_SAMPLE_RATE)
-      clip_bounds = np.array((x[clip[0]], x[clip[1]])) * AUDIO_SAMPLE_RATE
-      sample_points = np.linspace(*clip_bounds, num=num_samples, endpoint=False)
-      segment.append(interp(sample_points))
-    segment = np.hstack(segment)
-    return segment
-  
-  # compress dual channel audio to mono for use in pitch corrected stretching
-  # pytsmod's wsola treats channels separately, so without this it sounds weird
-  if not no_pitch_correction:
-    audio_desc_arr_mono = np.mean(audio_desc_arr, axis=0)
-  
-  x,y = zip(*smooth_path)
-  for run in runs:
-    run_length_seconds = y[run[-1][1]] - y[run[0][0]]
-    if run_length_seconds < MIN_DURATION_TO_REPLACE_SECONDS:
-      continue
-    anchor_point_path_indices = [clip[0] for clip in run]
-    anchor_point_path_indices.append(run[-1][1])
-    anchor_points = (np.array((np.array(x)[anchor_point_path_indices],
-                               np.array(y)[anchor_point_path_indices])) * AUDIO_SAMPLE_RATE).astype(int)
-    slopes = np.diff(anchor_points[1]) / np.diff(anchor_points[0])
-    for clip_index, (clip, slope) in enumerate(zip(run, slopes)):
-      # only apply pitch correction if the difference would be noticeable
-      if no_pitch_correction or np.abs(1 - slope) <= JUST_NOTICEABLE_DIFF_IN_FREQ_RATIO:
-        stretched_audio = get_interped_segment([clip], audio_desc_arr_interp)
+  # yields matrices of pearson correlations indexed by the first window's start and
+  # the second window's offset from the first window
+  # the output matrix is truncated to the valid square with positive offsets
+  # if negative=True, it is truncated to the valid square with negative offsets
+  # subsequent yields are the adjacent square following the previously yielded one
+  def get_pearson_corrs_generator(input, negative, jumps, window_size=512):
+    # processing the entire vector at once is faster, but uses too much memory
+    # instead, parse the input vector in pieces with a recursive call
+    max_cached_chunks = 50
+    cut = max_cached_chunks * window_size
+    if input.shape[1] > (max_cached_chunks + 2) * 1.1 * window_size:
+      is_first_iter = True
+      while True:
+        output_start = 0 if is_first_iter else 1
+        is_last_iter = (input.shape[1] <= (max_cached_chunks + 2) * 1.1 * window_size)
+        output_end = None if is_last_iter else max_cached_chunks
+        input_end = None if is_last_iter else (cut + window_size)
+        yield from itertools.islice(get_pearson_corrs_generator(input[:,:input_end], negative, jumps),
+                                    output_start, output_end)
+        if is_last_iter:
+          return
+        input = input[:,cut-window_size:]
+        is_first_iter = False
+    if input.shape[1] < 3 * window_size - 1:
+      raise RuntimeError("Invalid state in Pearson generator.")
+    pearson_corrs = np.zeros((len(jumps), input.shape[1] - window_size + 1)) - np.inf
+    # calculate dot products of pairs of windows (i.e. autocorrelation)
+    # avoids redundant calculations by substituting differences in the cumulative sum of products
+    self_corr = np.sum(input.astype(np.float32)**2, axis=0)
+    corr_cumsum = np.cumsum(self_corr, dtype=np.float64)
+    corr_cumsum[window_size:] -= corr_cumsum[:-window_size]
+    window_rms = corr_cumsum[window_size-1:]
+    epsilon = 1e-4 * max(1, np.max(window_rms))
+    window_rms = np.sqrt(window_rms + epsilon)
+    for jump_index, jump in enumerate(jumps):
+      autocorrelation = np.sum(input[:,jump:].astype(np.float32) * input[:,:input.shape[1]-jump], axis=0)
+      autocorr_cumsum = np.cumsum(autocorrelation, dtype=np.float64)
+      autocorr_cumsum[window_size:] -= autocorr_cumsum[:-window_size]
+      if negative:
+        pearson_corrs[jump_index, jump:] = autocorr_cumsum[window_size-1:] + epsilon
+        pearson_corrs[jump_index, jump:] /= window_rms[:len(window_rms)-jump]
       else:
-        anchor_point_pair = anchor_points[:,clip_index:clip_index+2].copy()
-        # account for quirks of pytsmod's wsola anchor point implementation
-        anchor_point_pair[1][-1] -= 1
-        anchor_y_offset = anchor_point_pair[1][0]
-        anchor_point_pair[1,:] -= anchor_y_offset
-        stretched_audio = pytsmod.wsola(audio_desc_arr_mono, anchor_point_pair)
-      video_arr[:,slice(*anchor_points[1,clip_index:clip_index+2])] = stretched_audio
-
-# identify which segments of the replaced audio actually have the describer speaking
-# uses a Naive Bayes classifier smoothed with L1-Minimization to identify the describer
-def detect_describer(video_arr, video_spec, video_spec_raw, video_timings,
-                     smooth_path, detect_sensitivity, boost_sensitivity):
-  # retokenize the audio description, which has been stretched to match the video
-  audio_desc_spec_raw, audio_timings = tokenize_audio(video_arr)
-  audio_desc_spec = normalize_spec(audio_desc_spec_raw)
-  
-  # avoid boosting or training on mismatched segments, like those close to skips
-  # assumes matching segments all have the same, constant play rate
-  # could be modified to handle a multi-modal distribution of rates
-  aligned_audio_times, aligned_video_times = zip(*smooth_path)
-  interp = scipy.interpolate.interp1d(aligned_video_times, aligned_audio_times,
-                                      fill_value = 'extrapolate',
-                                      bounds_error = False, assume_sorted = True)
-  slopes = (interp(video_timings + 1e-5) - \
-            interp(video_timings - 1e-5)) / 2e-5
-  median_slope = np.median(slopes)
-  aligned_mask =      np.abs(slopes - median_slope) < MAX_RATE_RATIO_DIFF_ALIGN
-  well_aligned_mask = np.abs(slopes - median_slope) < MAX_RATE_RATIO_DIFF_BOOST
-  
-  # first pass identification by assuming poorly matched tokens are describer speech
-  # also assumes the describer doesn't speak very quietly
-  corrs = np.sum(audio_desc_spec * video_spec, axis=-1)
-  smooth_volume = nd.gaussian_filter(audio_desc_spec[:,0], sigma=1)
-  audio_desc_loud = smooth_volume > np.percentile(smooth_volume, 30)
-  speech_mask = (corrs < .2) * audio_desc_loud
-  
-  # normalize spectrogram coefficients along time axis to prep for conversion to PDFs
-  audio_desc_spec = normalize_spec(audio_desc_spec_raw, axes=(0,))
-  audio_desc_spec = np.clip(audio_desc_spec / 6., -1, 1)
-  video_spec = normalize_spec(video_spec_raw, axes=(0,))
-  video_spec = np.clip(video_spec / 6., -1, 1)
-  
-  # convert sampled features (e.g. spectrogram) to probability densities of each feature
-  # when given a spectrogram, finds the distributions of the MFC coefficients
-  def make_log_pdfs(arr):
-    resolution = 100
-    bins_per_spot = 4
-    num_bins = int(resolution * bins_per_spot)
-    uniform_prior_strength_per_spot = 1
-    uniform_prior_strength_per_bin = uniform_prior_strength_per_spot / float(bins_per_spot)
-    bin_range = (-1 - 1e-10, 1 + 1e-10)
-    get_hist = lambda x: np.histogram(x, bins=num_bins, range=bin_range)[0]
-    pdfs = np.apply_along_axis(get_hist, 1, arr.T)
-    pdfs = pdfs + uniform_prior_strength_per_bin
-    smooth = lambda x: nd.gaussian_filter(x, sigma=bins_per_spot)
-    pdfs = np.apply_along_axis(smooth, 1, pdfs)
-    pdfs = pdfs / np.sum(pdfs[0,:])
-    log_pdfs = np.log(pdfs)
-    bin_edges = np.histogram([], bins=num_bins, range=bin_range)[1]
-    return log_pdfs, bin_edges
-  
-  diff_spec = audio_desc_spec - video_spec
-  diff_spec = np.clip(diff_spec, -1, 1)
+        pearson_corrs[jump_index, :pearson_corrs.shape[1]-jump] = autocorr_cumsum[window_size-1:] + epsilon
+        pearson_corrs[jump_index, :pearson_corrs.shape[1]-jump] /= window_rms[jump:]
+    # divide by RMS of constituent windows to get Pearson correlations
+    pearson_corrs = pearson_corrs / window_rms[None,:]
+    pearson_corrs = pearson_corrs.T
+    for chunk_index in range(0, input.shape[1] // window_size):
+      yield pearson_corrs[chunk_index*window_size:(chunk_index+1)*window_size]
   
-  # Naive Bayes classifier to roughly estimate whether each token is describer speech
-  desc_log_pdfs, _ = make_log_pdfs(diff_spec[speech_mask * well_aligned_mask])
-  nondesc_log_pdfs, bin_edges = make_log_pdfs(diff_spec[(~speech_mask) * well_aligned_mask])
-  lratio_lookup = desc_log_pdfs - nondesc_log_pdfs
-  lratios = lratio_lookup[np.fromfunction(lambda i,j: j, diff_spec.shape, dtype=int),
-                          np.digitize(diff_spec, bin_edges, right=True)-1]
-  ratio_desc_to_nondesc = np.sum(speech_mask * well_aligned_mask) /\
-                         (np.sum((~speech_mask) * well_aligned_mask) + 1.)
-  relative_probs = np.sum(lratios, axis=1)
-  relative_probs /= np.std(relative_probs)
-  relative_probs -= np.mean(relative_probs)
-  
-  # L1-Minimization to smoothly identify audio descriptions using a linear program
-  # x is fit_err_pos, fit_err_neg, delta_fit_pos, delta_fit_neg
-  # fit_err[i] = relative_probs[i] - y_fit[i]
-  # delta_fit[i] = y_fit[i] - y_fit[i-1]
-  # this can be rewritten in terms of fit_err by re-arranging the 1st equation:
-  #   y_fit[i] = relative_probs[i] - fit_err[i]
-  # this gives:
-  #   delta_fit[i] = (relative_probs[i] - relative_probs[i-1]) -\
-  #                  (fit_err[i] - fit_err[i-1])
-  # the delta_fit variables can then be set using equality constraints
-  num_fit_points = len(relative_probs)
-  y_diffs = np.diff(relative_probs)
-  pos_err_cost_factor = MIN_DESC_DURATION / float(TIMESTEP_SIZE_SECONDS)
-  neg_err_cost_factor = MAX_GAP_IN_DESC_SEC / float(TIMESTEP_SIZE_SECONDS)
-  c = np.hstack([np.ones(num_fit_points) / pos_err_cost_factor,
-                 np.ones(num_fit_points) / neg_err_cost_factor,
-                 np.ones(num_fit_points - 1) / 2.,
-                 np.ones(num_fit_points - 1) / 2.])
-  fit_err_coeffs = scipy.sparse.diags([-np.ones(num_fit_points),
-                                        np.ones(num_fit_points)],
-                                      offsets=[0,1],
-                                      shape=(num_fit_points - 1, num_fit_points)).tocsc()
-  A_eq = scipy.sparse.hstack([ fit_err_coeffs,
-                              -fit_err_coeffs,
-                               scipy.sparse.eye(num_fit_points-1),
-                              -scipy.sparse.eye(num_fit_points-1)])
-  b_eq = y_diffs
-  fit = scipy.optimize.linprog(c, A_eq=A_eq, b_eq=b_eq, method='highs-ds')
-  # if dual simplex solver encounters numerical problems, retry with interior point solver
-  if not fit.success and fit.status == 4:
-    fit = scipy.optimize.linprog(c, A_eq=A_eq, b_eq=b_eq, method='highs-ipm')
-  if not fit.success:
-    print(fit)
-    raise RuntimeError("Describer Voice Detection L1-Min Optimization Failed!")
-  
-  # combine fit_err_pos and fit_err_neg
-  fit_err = fit.x[:num_fit_points] - fit.x[num_fit_points:2*num_fit_points]
-  
-  # subtract fit errors from nodes to retrieve the smoothed fit
-  smooth_desc_locations = relative_probs - fit_err
-  
-  # hard threshold to classify each token as describer speech or not
-  speech_mask = smooth_desc_locations > 1. - 1.5 * detect_sensitivity
-  speech_mask *= aligned_mask
-  
-  # a separate mask is created for describer volume boosting
-  # as losing the describer's voice entirely is usually worse than it just being quiet
-  # and imperfectly aligned segments may have descriptions, but shouldn't be boosted
-  boost_mask = smooth_desc_locations > 1. - 1.5 * boost_sensitivity
-  boost_mask *= well_aligned_mask
-  
-  # convert a token classification into a mask that can be applied directly to samples
-  # unlike the input, the output isn't a boolean array but an array of floats
-  def token_mask_to_sample_mask(token_mask):
-    description_timings = video_timings[1:-1][token_mask[1:-1]]
-    sample_mask = np.zeros(video_arr.shape[1], dtype=np.float32)
-    window_radius = int(AUDIO_SAMPLE_RATE * TIMESTEP_SIZE_SECONDS)
-    window_size_seconds = 2 * window_radius + 1
-    bump = scipy.signal.windows.hann(window_size_seconds)
-    for description_timing in description_timings:
-      window_center = int(description_timing * AUDIO_SAMPLE_RATE)
-      sample_mask[window_center-window_radius:window_center+window_radius+1] += bump
-    return sample_mask
-  
-  speech_sample_mask = token_mask_to_sample_mask(speech_mask)
-  boost_sample_mask = token_mask_to_sample_mask(boost_mask)
-  ad_timings = video_timings.copy()
-  ad_timings[~speech_mask] = np.inf
+  def stretch(input, output, window_size=512, max_drift=512*3):
+    drift_window_size = max_drift * 2 + 1
+    num_input_samples = input.shape[1]
+    num_output_samples = output.shape[1]
+    total_offset_samples = num_output_samples - num_input_samples
+    jumps = [506, 451, 284, 410, 480, 379, 308, 430, 265, 494]
+    # use all jumps when given unreachable or difficult to reach offsets (i.e. Frobenius coin problem)
+    # otherwise, skip most jumps to trade off a little performance for a lot of speed
+    if abs(total_offset_samples) < 10000:
+      if abs(total_offset_samples) > 1000:
+        jumps.extend([MIN_STRETCH_OFFSET + offset for offset in (2**np.arange(8))-1])
+      else:
+        jumps = range(MIN_STRETCH_OFFSET, window_size)
+    num_windows = (num_input_samples // window_size)
+    window_to_offset = lambda window_index: (total_offset_samples * \
+                                             min((num_windows - 1), max(0, window_index))) // (num_windows - 1)
+    # note the absolute value in the drift
+    # the following calculations also use the absolute value of the jumps
+    # their signs flip together, so this saves on casework in the code
+    # after the optimal route is determined, the sign of the jumps will be reintroduced
+    window_to_offset_diff = lambda window_index: abs(window_to_offset(window_index) - \
+                                                     window_to_offset(window_index - 1))
+    backpointers = np.zeros((num_windows, drift_window_size), dtype=np.int16)
+    best_jump_locations = np.zeros((num_windows, len(jumps)), dtype=np.int16)
+    cum_loss = np.zeros((3, drift_window_size)) + np.inf
+    cum_loss[1:, max_drift] = 0
+    last_offset_diff = 0
+    # if the output needs to be longer than the input, we need to jump backwards in the input
+    pearson_corrs_generator = get_pearson_corrs_generator(input, (total_offset_samples > 0), jumps)
+    for window_index in range(num_windows):
+      corrs = next(pearson_corrs_generator)
+      # for each jump distance, determine the best input index in the window to make that jump
+      best_jump_locations[window_index] = np.argmax(corrs, axis=0)
+      best_jump_losses = 1 - corrs[best_jump_locations[window_index], np.arange(corrs.shape[1])]
+      offset_diff = window_to_offset_diff(window_index)
+      offset_diff2 = offset_diff + last_offset_diff
+      offset_jump_losses = np.zeros((len(jumps)+1, drift_window_size)) + np.inf
+      # consider not jumping at all, copying the loss from the corresponding offset one window back
+      offset_jump_slice = slice(None, offset_jump_losses.shape[1] - offset_diff)
+      offset_jump_losses[0,offset_jump_slice] = cum_loss[(window_index-1)%3,offset_diff:]
+      for jump_index, jump in enumerate(jumps):
+        truncation_amount = offset_diff2 - jump
+        offset_jump_slice = slice(jump, drift_window_size - max(0, truncation_amount))
+        cum_loss_slice = slice(offset_diff2, drift_window_size + min(0, truncation_amount))
+        # consider jumping the given distance from two windows back
+        # a window is skipped when jumping to prevent overlapping crossfades
+        offset_jump_losses[jump_index+1, offset_jump_slice] = cum_loss[(window_index-2)%3, cum_loss_slice] + \
+                                                              best_jump_losses[jump_index]
+      best_jumps = np.argmin(offset_jump_losses, axis=0)
+      backpointers[window_index] = best_jumps
+      cum_loss[window_index%3] = offset_jump_losses[best_jumps, np.arange(offset_jump_losses.shape[1])]
+      last_offset_diff = offset_diff
+    drift = max_drift
+    best_jumps = []
+    skip_window = False
+    for window_index in range(num_windows - 1, -1, -1):
+      drift += window_to_offset_diff(window_index + 1)
+      if skip_window:
+        skip_window = False
+        continue
+      best_jump_index = backpointers[window_index, drift] - 1
+      if best_jump_index == -1:
+        continue
+      best_jump = jumps[best_jump_index]
+      jump_input_index = window_index * window_size + \
+                         best_jump_locations[window_index, best_jump_index].item()
+      drift -= best_jump
+      skip_window = True
+      best_jumps.append((jump_input_index, best_jump))
+    best_jumps = best_jumps[::-1]
+    best_jumps = np.array(best_jumps)
+    # reintroduce the sign of the jump distances
+    # if the output is longer, use backwards jumps in the input to duplicate samples
+    # if the output is shorter, use forwards jumps in the input to remove samples
+    if total_offset_samples > 0:
+      best_jumps[:,1] *= -1
+    jump_input_indices = best_jumps[:,0]
+    jump_distances = best_jumps[:,1]
+    # calculate starts and ends of segments that will be copied from input to output
+    input_starts = np.concatenate(([0], jump_input_indices + jump_distances))
+    input_ends = np.concatenate((jump_input_indices, [input.shape[1]]))
+    chunk_lengths = input_ends - input_starts
+    output_ends = np.cumsum(chunk_lengths)
+    output_starts = np.concatenate(([0], output_ends[:-1]))
+    bump = scipy.signal.windows.hann(2 * window_size + 1)
+    bump_head = bump[:window_size]
+    bump_tail = bump[window_size:-1]
+    output[:,:window_size] = input[:,:window_size]
+    for in_start, in_end, out_start, out_end in zip(input_starts, input_ends, output_starts, output_ends):
+      output[:,out_start:out_start+window_size] *= bump_tail
+      output[:,out_start:out_start+window_size] += input[:,in_start:in_start+window_size] * bump_head
+      output[:,out_start+window_size:out_end+window_size] = input[:,in_start+window_size:in_end+window_size]
   
-  return speech_sample_mask, boost_sample_mask, ad_timings
+  x = audio_desc_times
+  y = video_times
+  x_samples = (x * AUDIO_SAMPLE_RATE).astype(int)
+  y_samples = (y * AUDIO_SAMPLE_RATE).astype(int)
+  diff_x_samples = np.diff(x_samples)
+  diff_y_samples = np.diff(y_samples)
+  slopes = diff_x_samples / diff_y_samples
+  total_offset_samples = diff_y_samples - diff_x_samples
+  y_midpoint_samples = (y_samples[:-1] + y_samples[1:]) // 2
+  progress_update_interval = (video_arr.shape[1] // 100) + 1
+  last_progress_update = -1
+  for i in range(len(x) - 1):
+    if diff_y_samples[i] < (MIN_DURATION_TO_REPLACE_SECONDS * AUDIO_SAMPLE_RATE) or \
+       np.abs(1 - slopes[i]) > MAX_RATE_RATIO_DIFF_ALIGN:
+      continue
+    video_arr_slice = video_arr[:,slice(*y_samples[i:i+2])]
+    progress = int(y_midpoint_samples[i] // progress_update_interval)
+    if progress > last_progress_update:
+      last_progress_update = progress
+      print(f"  stretching audio:{progress:3d}%                        \r", end='')
+    # only apply pitch correction if the difference would be noticeable
+    if no_pitch_correction or np.abs(1 - slopes[i]) <= JUST_NOTICEABLE_DIFF_IN_FREQ_RATIO or \
+       abs(total_offset_samples[i]) < MIN_STRETCH_OFFSET:
+      # construct a stretched audio description waveform using the quadratic interpolator
+      sample_points = np.linspace(*x_samples[i:i+2], num=diff_y_samples[i], endpoint=False)
+      video_arr_slice[:] = audio_desc_arr_interp(sample_points)
+    else:
+      stretch(audio_desc_arr[:,slice(*x_samples[i:i+2])], video_arr_slice)
 
 # Convert piece-wise linear fit to ffmpeg expression for editing video frame timestamps
-def encode_fit_as_ffmpeg_expr(smooth_path, clips, video_offset, start_key_frame):
+def encode_fit_as_ffmpeg_expr(audio_desc_times, video_times, video_offset):
   # PTS is the input frame's presentation timestamp, which is when frames are displayed
   # TB is the timebase, which is how many seconds each unit of PTS corresponds to
   # the output value of the expression will be the frame's new PTS
   setts_cmd = ['TS']
-  start_skip = max(0, video_offset - start_key_frame)
-  if start_skip > 0:
-    # lossless cutting can only happen at key frames, so we cut the video before the audio starts
-    # but that means the video is behind the audio and needs to catch up by playing quicker
-    # catchup_spread is the ratio of time to spend catching up to the amount of catching up needed
-    catchup_spread = 1./CATCHUP_RATE
-    setts_cmd.append(f'+clip(TS-{start_key_frame},0,{start_skip*(1+catchup_spread)}/TB)*{-1./(1+catchup_spread)}')
-  elif video_offset < 0:
-    # if the audio starts before the video, stretch the first frame of the video back to meet it
-    setts_cmd.append(f'+clip(TS-{start_key_frame},0,{-video_offset/10000.}/TB)*10000')
   # each segment of the linear fit can be encoded as a single clip function
   setts_cmd.append('+(0')
-  for clip_start, clip_end in clips:
-    audio_desc_start, video_start = smooth_path[clip_start]
-    audio_desc_end, video_end = smooth_path[clip_end]
-    video_start -= start_key_frame
-    video_end -= start_key_frame
-    audio_desc_length = audio_desc_end - audio_desc_start
-    video_length = video_end - video_start
-    slope = audio_desc_length / video_length
-    setts_cmd.append(f'+clip(TS-{start_key_frame}-{video_start:.4f}/TB,0,{max(0,video_length):.4f}/TB)*{slope-1:.9f}')
+  x = audio_desc_times
+  y = video_times
+  diff_x = np.diff(x)
+  diff_y = np.diff(y)
+  slopes = diff_x / diff_y
+  for i in range(len(audio_desc_times) - 1):
+    setts_cmd.append(f'+clip(TS-{y[i]-video_offset:.4f}/TB,0,{max(0,diff_y[i]):.4f}/TB)*{slopes[i]-1:.9f}')
   setts_cmd.append(')')
   setts_cmd = ''.join(setts_cmd)
   return setts_cmd
@@ -759,75 +433,65 @@ def get_ffmpeg():
 def get_ffprobe():
   return static_ffmpeg.run._get_or_fetch_platform_executables_else_raise_no_lock()[1]
 
+def get_key_frame_data(video_file, time=None, entry='pts_time'):
+  interval = f'%+{max(60,time+40)}' if time != None else '%'
+  key_frames = ffmpeg.probe(video_file, cmd=get_ffprobe(), select_streams='V', show_frames=None, 
+                            skip_frame='nokey', read_intervals=interval,
+                            show_entries='frame='+entry)['frames']
+  return np.array([float(frame[entry]) for frame in key_frames if entry in frame])
+
+# finds the average timestamp of (i.e. midpoint between) the key frames on either side of input time
 def get_closest_key_frame_time(video_file, time):
-  if time <= 0:
-    return 0
-  key_frames = ffmpeg.probe(video_file, cmd=get_ffprobe(), select_streams='V',
-                            show_frames=None, skip_frame='nokey')['frames']
-  key_frame_times = np.array([float(frame['pts_time']) for frame in key_frames] + [0])
-  return np.max(key_frame_times[key_frame_times <= time])
+  key_frame_times = get_key_frame_data(video_file, time)
+  key_frame_times = key_frame_times if len(key_frame_times) > 0 else np.array([0])
+  prev_key_frame_times = key_frame_times[key_frame_times <= time]
+  prev_key_frame = np.max(prev_key_frame_times) if len(prev_key_frame_times) > 0 else time
+  next_key_frame_times = key_frame_times[key_frame_times > time]
+  next_key_frame = np.min(next_key_frame_times) if len(next_key_frame_times) > 0 else time
+  return (prev_key_frame + next_key_frame) / 2.
 
 # outputs a new media file with the replaced audio (which includes audio descriptions)
 def write_replaced_media_to_disk(output_filename, media_arr, video_file=None, audio_desc_file=None,
-                                 setts_cmd=None, start_key_frame=None):
-  if audio_desc_file is None:
-    media_input = ffmpeg.input('pipe:', format='s16le', acodec='pcm_s16le',
-                               ac=2, ar=AUDIO_SAMPLE_RATE)
-    if video_file is None or os.path.splitext(output_filename)[1][1:] in AUDIO_EXTENSIONS:
-      write_command = ffmpeg.output(media_input, output_filename, loglevel='fatal').overwrite_output()
+                                 setts_cmd=None, video_offset=None, after_start_key_frame=None):
+  # if a media array is given, stretch_audio is enabled and media_arr should be added to the video
+  if media_arr is not None:
+    media_input = ffmpeg.input('pipe:', format='s16le', acodec='pcm_s16le', ac=2, ar=AUDIO_SAMPLE_RATE)
+    # if no video file is given, the input "video" was an audio file and the output should be too
+    if video_file is None:
+      write_command = ffmpeg.output(media_input, output_filename, loglevel='error').overwrite_output()
     else:
-      original_video = ffmpeg.input(video_file)
+      original_video = ffmpeg.input(video_file, dn=None)
       # "-max_interleave_delta 0" is sometimes necessary to fix an .mkv bug that freezes audio/video:
       #   ffmpeg bug warning: [matroska @ 0000000002c814c0] Starting new cluster due to timestamp
       # more info about the bug and fix: https://reddit.com/r/ffmpeg/comments/efddfs/
       write_command = ffmpeg.output(media_input, original_video, output_filename,
                                     acodec='copy', vcodec='copy', scodec='copy',
-                                    max_interleave_delta='0', loglevel='fatal',
-                                    **{"c:a:0": "aac", "disposition:a:0": "default"}).overwrite_output()
-    ffmpeg_caller = write_command.run_async(pipe_stdin=True, cmd=get_ffmpeg())
-    ffmpeg_caller.stdin.write(media_arr.astype(np.int16).T.tobytes())
-    ffmpeg_caller.stdin.close()
-    ffmpeg_caller.wait()
+                                    max_interleave_delta='0', loglevel='error',
+                                    **{"c:a:0": "aac", "disposition:a:1": "original",
+                                       "metadata:s:a:1": "title=original",
+                                       "disposition:a:0": "default+visual_impaired+descriptions",
+                                       "metadata:s:a:0": "title=AD"}).overwrite_output()
+    run_async_ffmpeg_command(write_command, media_arr, f"write output file: {output_filename}")
   else:
-    media_input = ffmpeg.input(audio_desc_file)
-    audio_desc_streams = ffmpeg.probe(audio_desc_file, cmd=get_ffprobe(), select_streams='a',
-                                      show_entries='format=duration')['streams']
-    audio_desc_duration = max([float(stream['duration']) for stream in audio_desc_streams])
-    original_video = ffmpeg.input(video_file, an=None, ss=start_key_frame)
-    if os.path.splitext(output_filename)[1] == os.path.splitext(video_file)[1]:
-      # wav files don't have codecs compatible with most video containers, so we convert to aac
-      audio_codec = 'copy' if os.path.splitext(audio_desc_file)[1] != '.wav' else 'aac'
-      # flac audio may only have experimental support in some video containers (e.g. mp4)
-      standards = 'normal' if os.path.splitext(audio_desc_file)[1] != '.flac' else 'experimental'
-      write_command = ffmpeg.output(media_input, original_video, output_filename,
-                                    acodec=audio_codec, vcodec='copy', scodec='copy',
-                                    max_interleave_delta='0', loglevel='fatal', strict=standards,
-                                    **{'bsf:v': f'setts=ts=\'{setts_cmd}\'',
-                                       'bsf:s': f'setts=ts=\'{setts_cmd}\''}).overwrite_output()
-      write_command.run(cmd=get_ffmpeg())
-    else:
-      # work around for bug that sometimes breaks setts when output and input formats differ
-      # the trick is separating the input and output by piping from one ffmpeg process into another
-      # mkv files break if 'nut' is used, while other files break when 'matroska' is used
-      format = 'matroska' if os.path.splitext(output_filename)[1] == '.mkv' else 'nut'
-      write_command = ffmpeg.output(original_video, 'pipe:', format=format, vsync='passthrough',
-                                    c='copy', loglevel='fatal')
-      ffmpeg_caller = write_command.run_async(pipe_stdout=True, cmd=get_ffmpeg())
-      pipe_input = ffmpeg.input('pipe:', format=format, thread_queue_size='512')
-      write_command2 = ffmpeg.output(media_input, pipe_input, output_filename, c='copy',
-                                     max_interleave_delta='0', loglevel='fatal', vsync='passthrough',
-                                     **{'bsf:v': f'setts=ts=\'{setts_cmd}\'',
-                                        'bsf:s': f'setts=ts=\'{setts_cmd}\''}).overwrite_output()
-      ffmpeg_caller2 = write_command2.run_async(pipe_stdin=True, cmd=get_ffmpeg())
-      while True:
-        in_bytes = ffmpeg_caller.stdout.read(100000)
-        if not in_bytes:
-          break
-        ffmpeg_caller2.stdin.write(in_bytes)
-      ffmpeg_caller2.stdin.close()
-      ffmpeg_caller.wait()
-      ffmpeg_caller2.wait()
-
+    start_offset = video_offset - after_start_key_frame
+    media_input = ffmpeg.input(audio_desc_file, itsoffset=f'{max(0, start_offset):.6f}')
+    original_video = ffmpeg.input(video_file, an=None, ss=f'{after_start_key_frame:.6f}',
+                                  itsoffset=f'{max(0, -start_offset):.6f}', dn=None)
+    # wav files don't have codecs compatible with most video containers, so we convert to aac
+    audio_codec = 'copy' if os.path.splitext(audio_desc_file)[1] != '.wav' else 'aac'
+    # flac audio may only have experimental support in some video containers (e.g. mp4)
+    standards = 'normal' if os.path.splitext(audio_desc_file)[1] != '.flac' else 'experimental'
+    # add frag_keyframe flag to prevent some players from ignoring audio/video start offsets
+    # set both pts and dts simultaneously in video manually, as ts= does not do the same thing
+    write_command = ffmpeg.output(media_input, original_video, output_filename,
+                                  acodec=audio_codec, vcodec='copy', scodec='copy',
+                                  max_interleave_delta='0', loglevel='error',
+                                  strict=standards, movflags='frag_keyframe',
+                                  **{'bsf:v': f'setts=pts=\'{setts_cmd}\':dts=\'{setts_cmd}\'',
+                                     'bsf:s': f'setts=ts=\'{setts_cmd}\'',
+                                     "disposition:a:0": "default+visual_impaired+descriptions",
+                                     "metadata:s:a:0": "title=AD"}).overwrite_output()
+    run_ffmpeg_command(write_command, f"write output file: {output_filename}")
 
 # check whether static_ffmpeg has already installed ffmpeg and ffprobe
 def is_ffmpeg_installed():
@@ -835,15 +499,482 @@ def is_ffmpeg_installed():
   indicator_file = os.path.join(ffmpeg_dir, "installed.crumb")
   return os.path.exists(indicator_file)
 
+def get_energy(arr):
+  # downsample of 105, hann size 15, downsample by 2 gives 210 samples per second, ~65 halfwindows/second
+  decimation = 105
+  decimation2 = 2
+  arr_clip = arr[:,:(arr.shape[1] - (arr.shape[1] % decimation))].reshape(arr.shape[0], -1, decimation)
+  energy = np.einsum('ijk,ijk->j', arr_clip, arr_clip, dtype=np.float32) / (decimation * arr.shape[0])
+  hann_window = scipy.signal.windows.hann(15)[1:-1].astype(np.float32)
+  hann_window /= np.sum(hann_window)
+  energy_smooth = np.convolve(energy, hann_window, mode='same')
+  energy_smooth = np.log10(1 + energy_smooth) / 2.
+  return energy_smooth[::decimation2]
+
+def get_zero_crossings(arr):
+  xings = np.diff(np.signbit(arr), prepend=False, axis=-1)
+  xings_clip = xings[:,:(xings.shape[1] - (xings.shape[1] % 210))].reshape(xings.shape[0], -1, 210)
+  zero_crossings = np.sum(np.abs(xings_clip), axis=(0,2)).astype(np.float32)
+  if xings.shape[0] == 1:
+    zero_crossings *= 2
+  hann_window = scipy.signal.windows.hann(15)[1:-1].astype(np.float32)
+  hann_window = hann_window / np.sum(hann_window)
+  zero_crossings_smooth = np.convolve(zero_crossings, hann_window, mode='same')
+  return zero_crossings_smooth
+
+def downsample_blur(arr, downsample, blur):
+  hann_window = scipy.signal.windows.hann(downsample*blur+2)[1:-1].astype(np.float32)
+  hann_window = hann_window / np.sum(hann_window)
+  arr = arr[:len(arr)-(len(arr)%downsample)]
+  return sum((np.convolve(arr[i::downsample], hann_window[i::downsample],
+                          mode='same') for i in range(downsample)))
+
+def get_freq_bands(arr):
+  arr = np.mean(arr, axis=0) if arr.shape[0] > 1 else arr[0]
+  arr = arr[:len(arr)-(len(arr)%210)]
+  downsamples = [5, 7, 6]
+  decimation = 1
+  freq_bands = []
+  for downsample in downsamples:
+    if downsample == downsamples[-1]:
+      band_bottom = np.array(0).reshape(1)
+    else:
+      band_bottom = downsample_blur(arr, downsample, 3)
+    decimation *= downsample
+    arr = arr.reshape(-1, downsample)
+    band_energy = sum(((arr[:,i] - band_bottom) ** 2 for i in range(downsample)))
+    freq_band = downsample_blur(band_energy, (210 // decimation), 15) / 210
+    freq_band = np.log10(1 + freq_band) / 2.
+    freq_bands.append(freq_band)
+    arr = band_bottom
+  return freq_bands
+
+def align(video_features, audio_desc_features, video_energy, audio_desc_energy):
+  samples_per_node = 210 // TIMESTEPS_PER_SECOND
+  hann_window_unnormed = scipy.signal.windows.hann(2*samples_per_node+1)[1:-1]
+  hann_window = hann_window_unnormed / np.sum(hann_window_unnormed)
+  get_mean = lambda arr: np.convolve(hann_window, arr, mode='same')
+  get_uniform_norm = lambda arr: np.convolve(np.ones(hann_window.shape), arr ** 2, mode='valid') ** .5
+  def get_uniform_norms(features):
+    return [np.clip(get_uniform_norm(feature), .001, None) for feature in features]
+  
+  print("  memorizing video...        \r", end='')
+  video_features_mean_sub = [feature - get_mean(feature) for feature in video_features]
+  audio_desc_features_mean_sub = [feature - get_mean(feature) for feature in audio_desc_features]
+  video_uniform_norms = get_uniform_norms(video_features_mean_sub)
+  audio_desc_uniform_norms = get_uniform_norms(audio_desc_features_mean_sub)
+  
+  num_bins = 7
+  bin_spacing = 6
+  bins_width = (num_bins - 1) * bin_spacing + 1
+  bins_start = samples_per_node - 1 - (bins_width // 2)
+  bins_end = bins_start + bins_width
+  video_dicts = [defaultdict(set) for feature in video_features_mean_sub]
+  edges = np.array(np.meshgrid(*([np.arange(2)]*num_bins), indexing='ij')).reshape(num_bins,-1).T
+  bin_offsets = []
+  for edge in edges:
+    bin_offset = np.array(np.meshgrid(*[np.arange(x+1) for x in edge], indexing='ij'))
+    bin_offsets.append(np.dot(bin_offset.reshape(num_bins,-1)[::-1].T, 7**np.arange(num_bins)))
+  
+  for video_dict, feature, norm in zip(video_dicts, video_features_mean_sub, video_uniform_norms):
+    bins = np.hstack([feature[bins_start+i:-bins_end+i+1, None] for i in bin_spacing * np.arange(num_bins)])
+    bins /= norm[:,None]
+    bins = 8 * bins + 3.3
+    np.clip(bins, 0, 6, out=bins)
+    bin_offset_indices = np.dot(((bins % 1) > .6), 2**np.arange(num_bins))
+    bins = np.dot(np.floor(bins).astype(int), 7**np.arange(num_bins)).tolist()
+    not_quiet = (video_energy[:-len(hann_window)] > .5)
+    for i in np.arange(len(video_energy) - len(hann_window))[not_quiet].tolist()[::4]:
+      bin = bins[i]
+      for bin_offset in bin_offsets[bin_offset_indices[i]].tolist():
+        video_dict[bin + bin_offset].add(i)
+  
+  print("  matching audio...  \r", end='')
+  audio_desc_bins = []
+  audio_desc_bin_offset_indices = []
+  for feature, norm in zip(audio_desc_features_mean_sub, audio_desc_uniform_norms):
+    bins = np.hstack([feature[bins_start+i:-bins_end+i+1, None] for i in bin_spacing * np.arange(num_bins)])
+    bins /= norm[:,None]
+    bins = 8 * bins + 3.5
+    bins = np.floor(bins).astype(int)
+    np.clip(bins, 0, 6, out=bins)
+    audio_desc_bins.append(np.dot(bins, 7**np.arange(num_bins)).tolist())
+  
+  def pairwise_intersection(set1, set2, set3):
+    return (set1 & set2).union((set1 & set3), (set2 & set3))
+  def triwise_intersection(set1, set2, set3, set4, set5):
+    set123 = pairwise_intersection(set1, set2, set3)
+    return (set123 & set4) | (set123 & set5)
+  best_so_far = SortedList(key=lambda x:x[0])
+  best_so_far.add((-1,-1,0))
+  backpointers = {}
+  not_quiet = (audio_desc_energy[:-len(hann_window)] > .5)
+  for i in np.arange(len(audio_desc_energy) - len(hann_window))[not_quiet].tolist():
+    match_sets = [video_dict[bins[i]] for bins, video_dict in zip(audio_desc_bins, video_dicts)]
+    common = triwise_intersection(*match_sets)
+    match_points = []
+    for video_index in common:
+      prob = 1
+      for j in range(3):
+        corr = np.dot(audio_desc_features_mean_sub[j][i:i+2*samples_per_node-1],
+                      video_features_mean_sub[j][video_index:video_index+2*samples_per_node-1])
+        corr /= audio_desc_uniform_norms[j][i] * video_uniform_norms[j][video_index]
+        prob *= max(1e-8, (1 - corr))  # Naive Bayes probability
+      prob = prob ** 2.9  # empirically determined, ranges from 2.5-3.4
+      if prob > 1e-8:
+        continue
+      qual = min(50, (prob / 1e-12) ** (-1. / 3))  # remove Naive Bayes assumption
+      match_points.append((video_index, qual))
+    audio_desc_index = i
+    for video_index, qual in sorted(match_points):
+      cur_index = best_so_far.bisect_right((video_index,))
+      prev_video_index, prev_audio_desc_index, prev_cum_qual = best_so_far[cur_index-1]
+      cum_qual = prev_cum_qual + qual
+      while (cur_index < len(best_so_far)) and (best_so_far[cur_index][2] <= cum_qual):
+        del best_so_far[cur_index]
+      best_so_far.add((video_index, audio_desc_index, cum_qual))
+      backpointers[(video_index, audio_desc_index)] = (prev_video_index, prev_audio_desc_index)
+  del video_dicts
+  path = [best_so_far[-1][:2]]
+  while path[-1][:2] in backpointers:
+    # failsafe to prevent an infinite loop that should never happen anyways
+    if len(path) > 10**8:
+      raise RuntimeError("Infinite Loop Encountered!")
+    path.append(backpointers[path[-1][:2]])
+  path.pop()
+  path.reverse()
+  if len(path) < max(min(len(video_energy), len(audio_desc_energy)) / 500., 5 * 210):
+    raise RuntimeError("Alignment failed, are the input files mismatched?")
+  y, x = np.array(path).T
+  
+  half_hann_window = hann_window[:samples_per_node-1] / np.sum(hann_window[:samples_per_node-1])
+  half_samples_per_node = samples_per_node // 2
+  fit_delay = samples_per_node + half_samples_per_node - 2
+  diff_by = lambda arr, offset=half_samples_per_node: arr[offset:] - arr[:-offset]
+  def get_continuity_err(x, y, deriv=False):
+    x_smooth_future = np.convolve(x, half_hann_window, mode='valid')
+    y_smooth_future = np.convolve(y, half_hann_window, mode='valid')
+    slopes_future = diff_by(y_smooth_future) / diff_by(x_smooth_future)
+    offsets_future = y_smooth_future[:-half_samples_per_node] - \
+                     x_smooth_future[:-half_samples_per_node] * slopes_future
+    x_smooth_past = np.convolve(x, half_hann_window[::-1], mode='valid')
+    y_smooth_past = np.convolve(y, half_hann_window[::-1], mode='valid')
+    slopes_past = diff_by(y_smooth_past) / diff_by(x_smooth_past)
+    offsets_past = y_smooth_past[half_samples_per_node:] - \
+                   x_smooth_past[half_samples_per_node:] * slopes_past
+    continuity_err = np.full(len(x) - (1 if deriv else 0), np.inf)
+    fit_delay_offset = fit_delay - (1 if deriv else 0)
+    continuity_err[:-fit_delay_offset] = np.abs(slopes_future * x[:-fit_delay] + \
+                                                offsets_future - y[:-fit_delay])
+    continuity_err[fit_delay_offset:] = np.minimum(continuity_err[fit_delay_offset:],
+                                                   np.abs(slopes_past * x[fit_delay:] + \
+                                                          offsets_past - y[fit_delay:]))
+    return continuity_err
+  
+  print("  refining match: pass 1 of 2...\r", end='')
+  continuity_err = get_continuity_err(x, y)
+  errs = (continuity_err < 3)
+  x = x[errs]
+  y = y[errs]
+  
+  audio_desc_features_scaled = []
+  video_features_scaled = []
+  for video_feature, audio_desc_feature in zip(video_features, audio_desc_features):
+    audio_desc_feature_std = np.std(audio_desc_feature)
+    scale_factor = np.linalg.lstsq(video_feature[y][:,None], audio_desc_feature[x], rcond=None)[0]
+    audio_desc_features_scaled.append(audio_desc_feature / audio_desc_feature_std)
+    video_features_scaled.append(video_feature * scale_factor / audio_desc_feature_std)
+  audio_desc_features_scaled = np.array(list(zip(*(audio_desc_features_scaled[:3]))))
+  video_features_scaled = np.array(list(zip(*(video_features_scaled[:3]))))
+  
+  smooth_x = get_mean(x)
+  smooth_y = get_mean(y)
+  slopes = np.diff(smooth_y) / np.diff(smooth_x)
+  offsets = smooth_y[:-1] - smooth_x[:-1] * slopes
+  err_y = slopes * x[:-1] + offsets - y[:-1]
+  compressed_x, compressed_y = [], []
+  def extend_all(index, compress=False, num=70):
+    compressed_x.extend([np.mean(x[index:index+num])] if compress else x[index:index+num])
+    compressed_y.extend([np.mean(y[index:index+num])] if compress else y[index:index+num])
+  extend_all(0, num=10)
+  for i in range(10, len(x) - 80, 70):
+    extend_all(i, compress=np.all(np.abs(err_y[i:i+70]) < 3))
+  extend_all(i+70)
+  
+  x = compressed_x
+  y = compressed_y
+  
+  match_dict = defaultdict(list)
+  x_unique = [-1]
+  for audio_desc_index, video_index in zip(x, y):
+    match_dict[audio_desc_index].append(video_index)
+    if audio_desc_index != x_unique[-1]:
+      x_unique.append(audio_desc_index)
+  x = np.array(x_unique[1:])
+  y = np.array([np.mean(match_dict[audio_desc_index]) for audio_desc_index in x])
+  
+  # L1-Minimization to solve the alignment problem using a linear program
+  # the absolute value functions needed for "absolute error" can be represented
+  # in a linear program by splitting variables into positive and negative pieces
+  # and constraining each to be positive (done by default in scipy's linprog)
+  num_fit_points = len(x)
+  x_diffs = np.diff(x)
+  y_diffs = np.diff(y)
+  jump_cost_base = 10.
+  jump_costs = np.full(num_fit_points - 1, jump_cost_base)
+  continuity_err = get_continuity_err(x, y, deriv=True)
+  jump_costs /= np.maximum(1, np.sqrt(continuity_err / 3.))
+  rate_change_jump_costs = np.full(num_fit_points - 1, .001)
+  rate_change_costs = np.full(num_fit_points - 2, jump_cost_base * 4000)
+  shot_noise_costs = np.full(num_fit_points, .01)
+  shot_noise_jump_costs = np.full(num_fit_points - 1, 3)
+  shot_noise_bound = 2.
+  c = np.hstack([np.ones(2 * num_fit_points),
+                 jump_costs,
+                 jump_costs,
+                 shot_noise_costs,
+                 shot_noise_costs,
+                 shot_noise_jump_costs,
+                 shot_noise_jump_costs,
+                 rate_change_jump_costs,
+                 rate_change_jump_costs,
+                 rate_change_costs,
+                 rate_change_costs,
+                 [0,]])
+  fit_err_coeffs = scipy.sparse.diags([-1. / x_diffs,
+                                        1. / x_diffs],
+                                      offsets=[0,1],
+                                      shape=(num_fit_points - 1, num_fit_points)).tocsc()
+  jump_coeffs = scipy.sparse.diags([ 1. / x_diffs],
+                                   offsets=[0],
+                                   shape=(num_fit_points - 1, num_fit_points - 1)).tocsc()
+  A_eq1 = scipy.sparse.hstack([ fit_err_coeffs,
+                               -fit_err_coeffs,
+                                jump_coeffs,
+                               -jump_coeffs,
+                                scipy.sparse.csc_matrix((num_fit_points - 1, 2 * num_fit_points)),
+                                jump_coeffs,
+                               -jump_coeffs,
+                                jump_coeffs,
+                               -jump_coeffs,
+                                scipy.sparse.csc_matrix((num_fit_points - 1, 2 * num_fit_points - 4)),
+                                np.ones((num_fit_points - 1, 1))])
+  A_eq2 = scipy.sparse.hstack([ scipy.sparse.csc_matrix((num_fit_points - 1, 4 * num_fit_points - 2)),
+                                scipy.sparse.diags([-1., 1.], offsets=[0, 1],
+                                                   shape=(num_fit_points - 1, num_fit_points)).tocsc(),
+                                scipy.sparse.diags([1., -1.], offsets=[0, 1],
+                                                   shape=(num_fit_points - 1, num_fit_points)).tocsc(),
+                               -scipy.sparse.eye(num_fit_points - 1),
+                                scipy.sparse.eye(num_fit_points - 1),
+                                scipy.sparse.csc_matrix((num_fit_points - 1, 4 * num_fit_points - 6)),
+                                scipy.sparse.csc_matrix((num_fit_points - 1, 1))])
+  slope_change_coeffs = scipy.sparse.diags([-1. / x_diffs[:-1],
+                                             1. / x_diffs[1:]],
+                                           offsets=[0,1],
+                                           shape=(num_fit_points - 2, num_fit_points - 1)).tocsc()
+  A_eq3 = scipy.sparse.hstack([scipy.sparse.csc_matrix((num_fit_points - 2, 8 * num_fit_points - 4)),
+                               slope_change_coeffs,
+                               -slope_change_coeffs,
+                               -scipy.sparse.eye(num_fit_points - 2),
+                               scipy.sparse.eye(num_fit_points - 2),
+                               scipy.sparse.csc_matrix((num_fit_points - 2, 1))])
+  A_eq = scipy.sparse.vstack([A_eq1, A_eq2, A_eq3])
+  b_eq = y_diffs / x_diffs
+  b_eq = np.hstack((b_eq, np.zeros(2 * num_fit_points - 3)))
+  bounds = [[0, None]] * (4 * num_fit_points - 2) + \
+           [[0, shot_noise_bound]] * (2 * num_fit_points) + \
+           [[0, None]] * (6 * num_fit_points - 8) + \
+           [[None, None]]
+  fit = scipy.optimize.linprog(c, A_eq=A_eq, b_eq=b_eq, bounds=bounds, method='highs-ds')
+  # if dual simplex solver encounters numerical problems, retry with interior point solver
+  if not fit.success and fit.status == 4:
+    fit = scipy.optimize.linprog(c, A_eq=A_eq, b_eq=b_eq, bounds=bounds, method='highs-ipm')
+  if not fit.success:
+    print(fit)
+    raise RuntimeError("Smooth Alignment L1-Min Optimization Failed!")
+  
+  # combine positive and negative components of variables
+  fit_err = fit.x[                  :  num_fit_points  ] - \
+            fit.x[  num_fit_points  :2*num_fit_points  ]
+  slope_jumps = fit.x[8*num_fit_points-4: 9*num_fit_points-5] - \
+                fit.x[9*num_fit_points-5:10*num_fit_points-6]
+  median_slope = fit.x[-1]
+  slopes = median_slope + (slope_jumps / x_diffs)
+  
+  # subtract fit errors from nodes to retrieve the smooth fit's coordinates
+  smooth_path = [(x, y) for x,y in zip(x, y - fit_err)]
+  
+  print("  refining match: pass 2 of 2...\r", end='')
+  slopes_plus_ends = np.hstack((slopes[:1], slopes, slopes[-1:]))
+  extensions = []
+  extend_radius = 210 * 30  # +/- 30 seconds
+  video_interp = scipy.interpolate.make_interp_spline(np.arange(len(video_features_scaled)),
+                                                      video_features_scaled, k=1)
+  colinear_dict = defaultdict(list)
+  for i, (x, y) in enumerate(smooth_path):
+    for slope in slopes_plus_ends[i:i+2]:
+      if (slope < .1) or (slope > 10):
+        continue
+      offset = y - slope * x
+      colinear_dict[(round(slope, 6), int(round(offset, 0)))].append((x, y))
+  line_clusters = []
+  added_keys = set()
+  for (slope, offset), indices in sorted(colinear_dict.items(), key=lambda x: -len(x[1])):
+    if (slope, offset) in added_keys:
+      continue
+    line_clusters.append(indices)
+    added_keys.add((slope, offset))
+    del colinear_dict[(slope, offset)]
+    for (slope2, offset2), indices2 in list(colinear_dict.items()):
+      if (abs(indices2[ 0][1] - (indices2[ 0][0] * slope + offset)) < 3) and \
+         (abs(indices2[-1][1] - (indices2[-1][0] * slope + offset)) < 3):
+        line_clusters[-1].extend(colinear_dict[(slope2, offset2)])
+        added_keys.add((slope2, offset2))
+        del colinear_dict[(slope2, offset2)]
+  line_clusters = [sorted(cluster) for cluster in line_clusters]
+  line_clusters = [x for x in line_clusters if (abs(x[0][0] - x[-1][0]) > 10) and len(x) > 5]
+  
+  for i, cluster in enumerate(line_clusters):
+    x, y = np.array(cluster).T
+    linear_fit = np.linalg.lstsq(np.hstack((np.ones((len(x), 1)), x[:, None])), y, rcond=None)[0]
+    line_clusters[i] = (x, linear_fit[0], linear_fit[1])
+  
+  def get_x_limits(x, offset, slope, extend_horiz=extend_radius, buffer_vert=4):
+    limits = (max(int(x[0])  - extend_horiz, 0),
+              min(int(x[-1]) + extend_horiz, len(audio_desc_features_scaled) - 1))
+    limits = (max(limits[0], int(np.ceil((buffer_vert - offset) / slope))),
+              min(limits[1], int(np.floor((len(video_features_scaled) - buffer_vert - offset) / slope))))
+    return limits
+  def get_audio_video_matches(limits, slope, offset):
+    x = np.arange(*limits)
+    y = slope * x + offset
+    audio_match = audio_desc_features_scaled[slice(*limits)]
+    video_match = video_interp(y)
+    return x, y, audio_match, video_match
+  
+  audio_desc_max_energy = np.max(audio_desc_features_scaled[:,0])
+  video_max_energy = np.max(video_features_scaled[:,0])
+  points = [[] for i in range(len(audio_desc_features_scaled))]
+  seen_points = set()
+  for cluster_index, (x, offset, slope) in enumerate(line_clusters):
+    limits = get_x_limits(x, offset, slope, extend_horiz=0)
+    if limits[1] < limits[0] + 5:
+      continue
+    if limits[1] > limits[0] + 100:
+      x, y, audio_match, video_match = get_audio_video_matches(limits, slope, offset)
+      video_match_err = audio_match[1:-1] - video_match[1:-1]
+      valid_matches = np.mean(video_match_err, axis=-1) < 0.1
+      if np.count_nonzero(valid_matches) > 50:
+        video_match_diff = (video_match[2:] - video_match[:-2]) / 2.
+        video_match_err = video_match_err[valid_matches]
+        video_match_diff = video_match_diff[valid_matches]
+        x_valid = x[1:-1][valid_matches][:,None]
+        A = video_match_diff.reshape(-1,1)
+        linear_fit, residual, _, _ = np.linalg.lstsq(A, video_match_err.flat, rcond=None)
+        explained_err_ratio = 1 - (residual / np.sum(video_match_err ** 2))
+        stds_above_noise_mean = np.sqrt(explained_err_ratio * np.prod(video_match_err.shape)) - 1.
+        if stds_above_noise_mean > 8 and abs(linear_fit[0]) < 2:
+          offset += linear_fit[0]
+    limits = get_x_limits(x, offset, slope)
+    x, y, audio_match, video_match = get_audio_video_matches(limits, slope, offset)
+    quals = np.sum(-.5 - np.log10(1e-4 + np.abs(audio_match - video_match)), axis=1)
+    quals *= np.clip(video_match[:,0] + 2.5 - video_max_energy, 0, 1)
+    quals += np.clip(audio_match[:,0] + 2.5 - audio_desc_max_energy, 0, 1) * .1
+    energy_diffs = audio_match[:,0] - video_match[:,0]
+    for i, j, qual in zip(x.tolist(), y.tolist(), quals.tolist()):
+      point = (i, int(j))
+      if point not in seen_points:
+        seen_points.add(point)
+        points[i].append((j, cluster_index, qual))
+  del seen_points
+  points = [sorted(point) for point in points]
+  
+  best_so_far = SortedList(key=lambda x:x[0])
+  best_so_far.add((0, 0, -1, 0, 0))  # video_index, audio_desc_index, cluster_index, qual, cum_qual
+  clusters_best_so_far = [(0, 0, 0, -1000) for cluster in line_clusters]
+  backpointers = {}
+  prev_cache = np.full((len(video_features_scaled), 5), -np.inf)
+  prev_cache[0] = (0, 0, -1, 0, 0)  # video_index, audio_desc_index, cluster_index, qual, cum_qual
+  reversed_min_points = [min(x)[0] if len(x) > 0 else np.inf for x in points[::-1]]
+  forward_min = list(itertools.accumulate(reversed_min_points, min))[::-1]
+  for i in range(len(audio_desc_features_scaled)):
+    for j, cluster_index, qual in points[i]:
+      cur_index = best_so_far.bisect_right((j,))
+      prev_j, prev_i, prev_cluster_index, prev_qual, best_prev_cum_qual = best_so_far[cur_index-1]
+      cluster_last = clusters_best_so_far[cluster_index]
+      if cluster_last[3] >= best_prev_cum_qual:
+        prev_j, prev_i, prev_qual, best_prev_cum_qual = cluster_last
+        prev_cluster_index = cluster_index
+      for prev_j_temp in range(max(0, int(j) - 2), int(j) + 1):
+        prev_node = prev_cache[prev_j_temp].tolist()
+        if cluster_index != prev_node[2]:
+          prev_node[4] -= 100 + 100 * ((j - prev_node[0]) - (i - prev_node[1])) ** 2
+        if prev_node[1] >= (i - 2) and \
+           prev_node[0] <= j and \
+           prev_node[4] >= best_prev_cum_qual:
+          prev_j, prev_i, prev_cluster_index, prev_qual, best_prev_cum_qual = prev_node
+      cum_qual = best_prev_cum_qual + qual
+      prev_cache[int(j)] = (j, i, cluster_index, qual, cum_qual)
+      cum_qual_jump = cum_qual - 1000
+      if best_so_far[cur_index-1][4] < cum_qual_jump:
+        while (cur_index < len(best_so_far)) and (best_so_far[cur_index][4] <= cum_qual_jump):
+          del best_so_far[cur_index]
+        best_so_far.add((j, i, cluster_index, qual, cum_qual_jump))
+      if forward_min[i] == j and cur_index > 1:
+        del best_so_far[:cur_index-1]
+      cum_qual_cluster_jump = cum_qual - 50
+      if cluster_last[3] < cum_qual_cluster_jump:
+        clusters_best_so_far[cluster_index] = (j, i, qual, cum_qual_cluster_jump)
+      backpointers[(j, i)] = (prev_j, prev_i, prev_cluster_index, prev_qual, best_prev_cum_qual)
+  path = [best_so_far[-1]]
+  while path[-1][:2] in backpointers:
+    path.append(backpointers[path[-1][:2]])
+  path.pop()
+  path.reverse()
+  path = np.array(path)
+  y, x, cluster_indices, quals, cum_quals = path.T
+  
+  nondescription = ((quals == 0) | (quals > .3))
+  similarity_ratio_x = float(len(set(x[nondescription]))) / len(audio_desc_features_scaled)
+  similarity_ratio_y = float(len(set(y[nondescription]))) / len(video_features_scaled)
+  similarity_percent = 100 * max(similarity_ratio_x, similarity_ratio_y)
+  
+  nodes = []
+  if cluster_indices[0] == cluster_indices[1]:
+    nodes.append((x[0], y[0]))
+  for i in range(len(x) - 1):
+    if cluster_indices[i] != cluster_indices[i+1]:
+      nodes.append((x[i] - .1, y[i] - .1))
+      nodes.append((x[i+1] + .1, y[i+1] + .1))
+  if cluster_indices[-2] == cluster_indices[-1]:
+    nodes.append((x[-1], y[-1]))
+  x, y = np.array(nodes).T / 210.
+  
+  if (x[1] - x[0]) > 2:
+    slope_start = (y[1] - y[0]) / (x[1] - x[0])
+    x[0] = 0
+    y[0] = y[1] - (x[1] * slope_start)
+    if y[0] < 0:
+      x[0] = x[1] - (y[1] / slope_start)
+      y[0] = 0
+  if (x[-1] - x[-2]) > 2:
+    slope_end = (y[-1] - y[-2]) / (x[-1] - x[-2])
+    x[-1] = ((len(audio_desc_energy) - 1) / 210.)
+    y[-1] = y[-2] + ((x[-1] - x[-2]) * slope_end)
+    if y[-1] > ((len(video_energy) - 1) / 210.):
+      y[-1] = ((len(video_energy) - 1) / 210.)
+      x[-1] = x[-2] + ((y[-1] - y[-2]) / slope_end)
+  
+  path[:,:2] /= 210.
+  return x, y, similarity_percent, path, median_slope
+
 # combines videos with matching audio files (e.g. audio descriptions)
 # this is the main function of this script, it calls the other functions in order
-def combine(video, audio, smoothness=50, stretch_audio=False, keep_non_ad=False,
-            boost=0, ad_detect_sensitivity=.6, boost_sensitivity=.4, yes=False,
-            prepend="ad_", no_pitch_correction=False, output_dir=default_output_dir,
-            alignment_dir=default_alignment_dir, extension="copy", display_func=None):
-  video_files, video_file_types = get_sorted_filenames(video, VIDEO_EXTENSIONS, AUDIO_EXTENSIONS)
+def combine(video, audio, stretch_audio=False, yes=False, prepend="ad_", no_pitch_correction=False,
+            output_dir=default_output_dir, alignment_dir=default_alignment_dir):
+  video_files, has_audio_extensions = get_sorted_filenames(video, VIDEO_EXTENSIONS, AUDIO_EXTENSIONS)
   
-  if yes == False and sum(video_file_types) > 0:
+  if yes == False and sum(has_audio_extensions) > 0:
     print("")
     print("One or more audio files found in video input. Was this intentional?")
     print("If not, press ctrl+c to kill this script.")
@@ -856,16 +987,16 @@ def combine(video, audio, smoothness=50, stretch_audio=False, keep_non_ad=False,
                  f"The audio path has {len(audio_desc_files)} files"]
     raise RuntimeError("\n".join(error_msg))
   
-  display("", display_func)
-  ensure_folders_exist([output_dir], display_func)
+  print("")
+  ensure_folders_exist([output_dir])
   if PLOT_ALIGNMENT_TO_FILE:
-    ensure_folders_exist([alignment_dir], display_func)
+    ensure_folders_exist([alignment_dir])
   
-  display("", display_func)
+  print("")
   for (video_file, audio_desc_file) in zip(video_files, audio_desc_files):
-    display(os.path.split(video_file)[1], display_func)
-    display(os.path.split(audio_desc_file)[1], display_func)
-    display("", display_func)
+    print(os.path.split(video_file)[1])
+    print(os.path.split(audio_desc_file)[1])
+    print("")
   if yes == False:
     print("Are the above input file pairings correct?")
     print("If not, press ctrl+c to kill this script.")
@@ -874,414 +1005,728 @@ def combine(video, audio, smoothness=50, stretch_audio=False, keep_non_ad=False,
   
   # if ffmpeg isn't installed, install it
   if not is_ffmpeg_installed():
-    display("Downloading and installing ffmpeg (media editor, 50 MB download)...", display_func)
+    print("Downloading and installing ffmpeg (media editor, 50 MB download)...")
     get_ffmpeg()
     if not is_ffmpeg_installed():
       RuntimeError("Failed to install ffmpeg.")
-    display("Successfully installed ffmpeg.", display_func)
+    print("Successfully installed ffmpeg.")
   
-  display("Processing files:", display_func)
+  print("Processing files:")
   
-  for (video_file, audio_desc_file, video_filetype) in zip(video_files, audio_desc_files,
-                                                           video_file_types):
-    # Default is to use the input video's extension for the output video
-    if extension is None or extension in ["", "copy"]:
-      ext = os.path.splitext(video_file)[1]
-    else:
-      # add a dot to the extension if it's missing
-      ext = ('' if extension[0] == '.' else '.') + extension
-    output_filename = prepend + os.path.splitext(os.path.split(video_file)[1])[0] + ext
+  for (video_file, audio_desc_file, has_audio_extension) in zip(video_files, audio_desc_files,
+                                                           has_audio_extensions):
+    # Output filename (and extension) is the same as input, except the prepend and directory
+    output_filename = prepend + os.path.split(video_file)[1]
     output_filename = os.path.join(output_dir, output_filename)
-    display(f" {output_filename}", display_func)
+    print(f" {output_filename}")
     
-    if os.path.exists(output_filename) and os.path.getsize(output_filename) > 0:
-      display("   output file already exists, skipping...", display_func)
+    if (not stretch_audio) & has_audio_extension:
+      raise RuntimeError("Argument --stretch_audio is required when both inputs are audio files.")
+    
+    if os.path.exists(output_filename) and os.path.getsize(output_filename) > 1e5:
+      print("   output file already exists, skipping...")
       continue
     
     # print warning if output file's full path is longer than Windows MAX_PATH (260)
     full_output_filename = os.path.abspath(output_filename)
     if IS_RUNNING_WINDOWS and len(full_output_filename) >= 260:
-      display("   WARNING: very long output path, ffmpeg may fail...", display_func)
+      print("  WARNING: very long output path, ffmpeg may fail...")
     
-    video_arr = parse_audio_from_file(video_file)
-    audio_desc_arr = parse_audio_from_file(audio_desc_file)
-    video_spec_raw, video_timings = tokenize_audio(video_arr)
-    video_spec = normalize_spec(video_spec_raw)
-    audio_desc_spec_raw, audio_desc_timings = tokenize_audio_dither(audio_desc_arr, video_timings)
-    audio_desc_spec = normalize_spec(audio_desc_spec_raw)
+    num_channels = 2 if stretch_audio else 1
+    print("  reading video file...\r", end='')
+    video_arr = parse_audio_from_file(video_file, num_channels)
     
-    # rescale RMS intensity of audio to match video
-    audio_desc_arr *= (np.std(video_arr) / np.std(audio_desc_arr))
+    print("  computing video features... \r", end='')
+    video_energy = get_energy(video_arr)
+    video_zero_crossings = get_zero_crossings(video_arr)
+    video_freq_bands = get_freq_bands(video_arr)
+    video_features = [video_energy, video_zero_crossings] + video_freq_bands
     
-    path, quals = rough_align(video_spec, audio_desc_spec, video_timings, audio_desc_timings)
+    if not stretch_audio:
+      del video_arr
     
-    similarity_ratio = float(len(quals)) / max(video_spec.shape[0], audio_desc_spec.shape[0])
-    similarity_percent = min(100, 100 * similarity_ratio)
-    if similarity_percent < 10:
-      display(f"   WARNING: similarity {similarity_percent:.1f}%, likely mismatched files", display_func)
-    if similarity_percent > 90:
-      display(f"   WARNING: similarity {similarity_percent:.1f}%, likely undescribed media", display_func)
+    print("  reading audio file...       \r", end='')
+    audio_desc_arr = parse_audio_from_file(audio_desc_file, num_channels)
+    
+    print("  computing audio features...\r", end='')
+    audio_desc_energy = get_energy(audio_desc_arr)
+    audio_desc_zero_crossings = get_zero_crossings(audio_desc_arr)
+    audio_desc_freq_bands = get_freq_bands(audio_desc_arr)
+    audio_desc_features = [audio_desc_energy, audio_desc_zero_crossings] + audio_desc_freq_bands
+    
+    if not stretch_audio:
+      del audio_desc_arr
     
-    smooth_path, runs, bad_clips, clips = smooth_align(path, quals, smoothness)
+    outputs = align(video_features, audio_desc_features, video_energy, audio_desc_energy)
+    audio_desc_times, video_times, similarity_percent, path, median_slope = outputs
     
-    cap_synced_end_points(smooth_path, video_arr, audio_desc_arr)
+    del video_energy, video_zero_crossings, video_freq_bands, video_features
+    del audio_desc_energy, audio_desc_zero_crossings, audio_desc_freq_bands, audio_desc_features
+    
+    if similarity_percent < 20:
+      print(f"  WARNING: similarity {similarity_percent:.1f}%, likely mismatched files")
+    if similarity_percent > 90:
+      print(f"  WARNING: similarity {similarity_percent:.1f}%, likely undescribed media")
     
-    ad_timings = None
     if stretch_audio:
-      if keep_non_ad:
-        video_arr_original = video_arr.copy()
+      # lower memory usage version of np.std for large arrays
+      def low_ram_std(arr):
+        avg = np.mean(arr, dtype=np.float64)
+        return np.sqrt(np.einsum('ij,ij->i', arr, arr, dtype=np.float64)/np.prod(arr.shape) - (avg**2))
       
-      replace_aligned_segments(video_arr, audio_desc_arr, smooth_path, runs, no_pitch_correction)
-      del audio_desc_arr
+      # rescale RMS intensity of audio to match video
+      audio_desc_arr *= (low_ram_std(video_arr) / low_ram_std(audio_desc_arr))[:, None]
       
-      if keep_non_ad or boost != 0:
-        outputs = detect_describer(video_arr, video_spec, video_spec_raw, video_timings,
-                                   smooth_path, ad_detect_sensitivity, boost_sensitivity)
-        speech_sample_mask, boost_sample_mask, ad_timings = outputs
-      if keep_non_ad:
-        video_arr *= speech_sample_mask
-        video_arr += video_arr_original * (1 - speech_sample_mask)
-        del video_arr_original
-        del speech_sample_mask
-      else:
-        ad_timings = None
-      if boost != 0:
-        video_arr = video_arr * (1. + (10**(boost / 10.) - 1.) * boost_sample_mask)
-        del boost_sample_mask
+      replace_aligned_segments(video_arr, audio_desc_arr, audio_desc_times, video_times, no_pitch_correction)
+      del audio_desc_arr
       
-      # prevent peaking by rescaling to within +/- 16,382
+      # prevent peaking by rescaling to within +/- 32,766
       video_arr *= (2**15 - 2.) / np.max(np.abs(video_arr))
       
-      if video_filetype == 0:
-        write_replaced_media_to_disk(output_filename, video_arr, video_file)
-      else:
-        write_replaced_media_to_disk(output_filename, video_arr)
+      print("  processing output file...                   \r", end='')
+      write_replaced_media_to_disk(output_filename, video_arr, None if has_audio_extension else video_file)
+      del video_arr
     else:
-      if video_filetype == 1:
-        raise RuntimeError("Argument --stretch_audio is required when both inputs are audio files.")
-      if os.path.splitext(output_filename)[1][1:] in AUDIO_EXTENSIONS:
-        raise RuntimeError("Argument --stretch_audio is required when output file extension is an audio filetype.")
-      video_offset = np.diff(smooth_path[clips[0][0]])[0]
-      start_key_frame = get_closest_key_frame_time(video_file, video_offset)
-      setts_cmd = encode_fit_as_ffmpeg_expr(smooth_path, clips, video_offset, start_key_frame)
+      video_offset = video_times[0] - audio_desc_times[0]
+      # to make ffmpeg cut at the last keyframe before the audio starts, use a timestamp after it
+      after_start_key_frame = get_closest_key_frame_time(video_file, video_offset)
+      print("  processing output file...                   \r", end='')
+      setts_cmd = encode_fit_as_ffmpeg_expr(audio_desc_times, video_times, video_offset)
       write_replaced_media_to_disk(output_filename, None, video_file, audio_desc_file,
-                                   setts_cmd, start_key_frame)
+                                   setts_cmd, video_offset, after_start_key_frame)
     
-    del video_arr
     if PLOT_ALIGNMENT_TO_FILE:
       plot_filename_no_ext = os.path.join(alignment_dir, os.path.splitext(os.path.split(video_file)[1])[0])
-      plot_alignment(plot_filename_no_ext, path, smooth_path, quals,
-                     runs, bad_clips, ad_timings, similarity_percent)
-  display("All files processed.", display_func)
+      plot_alignment(plot_filename_no_ext, path, audio_desc_times, video_times, similarity_percent,
+                     median_slope, stretch_audio, no_pitch_correction)
+  print("All files processed.       ")
 
-def write_config_file(config_path, settings):
-  config = configparser.ConfigParser()
-  config.add_section('alignment')
-  config['alignment'] = {}
-  for key, value in settings.items():
-    config['alignment'][key] = str(value)
-  with open(config_path, 'w') as f:
-    config.write(f)
+if wx is not None:
+  def write_config_file(config_path, settings):
+    config = configparser.ConfigParser()
+    config.add_section('alignment')
+    config['alignment'] = {}
+    for key, value in settings.items():
+      config['alignment'][key] = str(value)
+    with open(config_path, 'w') as f:
+      config.write(f)
 
-def read_config_file(config_path: Path):
-  config = configparser.ConfigParser()
-  config.read(config_path)
-  settings = {'smoothness':           config.getfloat('alignment', 'smoothness', fallback=50),
-              'stretch_audio':        config.getboolean('alignment', 'stretch_audio', fallback=False),
-              'keep_non_ad':          config.getboolean('alignment', 'keep_non_ad', fallback=False),
-              'boost':                config.getfloat('alignment', 'boost', fallback=0),
-              'ad_detect_sensitivity':config.getfloat('alignment', 'ad_detect_sensitivity', fallback=.6),
-              'boost_sensitivity':    config.getfloat('alignment', 'boost_sensitivity', fallback=.4),
-              'prepend':              config.get('alignment', 'prepend', fallback='ad_'),
-              'no_pitch_correction':  config.getboolean('alignment', 'no_pitch_correction', fallback=False),
-              'output_dir':           config.get('alignment', 'output_dir', fallback=default_output_dir),
-              'alignment_dir':        config.get('alignment', 'alignment_dir', fallback=default_alignment_dir),
-              'extension':            config.get('alignment', 'extension', fallback='copy')}
-  if not config.has_section('alignment'):
-    write_config_file(config_path, settings)
-  return settings
+  def read_config_file(config_path: Path):
+    config = configparser.ConfigParser()
+    config.read(config_path)
+    settings = {'stretch_audio':       config.getboolean('alignment', 'stretch_audio', fallback=False),
+                'prepend':             config.get('alignment', 'prepend', fallback='ad_'),
+                'no_pitch_correction': config.getboolean('alignment', 'no_pitch_correction', fallback=False),
+                'output_dir':          config.get('alignment', 'output_dir', fallback=default_output_dir),
+                'alignment_dir':       config.get('alignment', 'alignment_dir', fallback=default_alignment_dir)}
+    if not config.has_section('alignment'):
+      write_config_file(config_path, settings)
+    return settings
+  
+  def set_tooltip(element, tip):
+    element.SetToolTip(tip)
+    # prevent tooltip from disappearing for 30 seconds
+    tooltip_object = element.GetToolTip()
+    if not tooltip_object is None:
+      tooltip_object.SetAutoPop(30000)
+  
+  class DialogSettings(wx.Dialog):
+    def __init__(self, parent, config_path, is_dark):
+      wx.Dialog.__init__(self, parent, title="Settings - describealign", size=wx.Size(450,330),
+                         style=wx.DEFAULT_DIALOG_STYLE|wx.TAB_TRAVERSAL)
+      # setting the GUI dialog's font causes all contained elements to inherit that font by default
+      self.SetFont(wx.Font(*gui_font))
+      self.SetBackgroundColour(gui_background_color_dark if is_dark else gui_background_color_light)
+      
+      self.text_header = wx.StaticText(self, label="Check tooltips (i.e. mouse-over text) for descriptions:")
+      
+      self.static_box_sizer_output = wx.StaticBoxSizer(wx.VERTICAL, self, "output_dir")
+      self.dir_picker_output = wx.DirPickerCtrl(self, message="Select a folder", name="output_dir")
+      set_tooltip(self.dir_picker_output, "Directory combined output media is saved to. " + \
+                                          "Default is \"videos_with_ad\"")
+      
+      self.static_box_sizer_alignment = wx.StaticBoxSizer(wx.VERTICAL, self, "alignment_dir")
+      self.dir_picker_alignment = wx.DirPickerCtrl(self, message="Select a folder", name="alignment_dir")
+      set_tooltip(self.dir_picker_alignment, "Directory alignment data and plots are saved to. " + \
+                                             "Default is \"alignment_plots\"")
+      
+      self.text_prepend = wx.StaticText(self, label="prepend:")
+      self.text_ctrl_prepend = wx.TextCtrl(self, name="prepend")
+      set_tooltip(self.text_ctrl_prepend, "Output file name prepend text. Default is \"ad_\"")
+      
+      panel_stretch_audio_no_pitch_correction = wx.Panel(self)
+      
+      self.checkbox_stretch_audio = wx.CheckBox(panel_stretch_audio_no_pitch_correction,
+                                                label="stretch_audio", name="stretch_audio")
+      set_tooltip(self.checkbox_stretch_audio, "Stretches the input audio to fit the input video. " + \
+                                               "Default is to stretch the video to fit the audio. " + \
+                                               "Keeps original video audio as secondary tracks. Slower " + \
+                                               "and uses more RAM when enabled, long videos may cause " + \
+                                               "paging or Out of Memory errors on low-RAM systems.")
+      self.checkbox_stretch_audio.Bind(wx.EVT_CHECKBOX, self.update_stretch_audio_subsettings)
+      
+      self.checkbox_no_pitch_correction = wx.CheckBox(panel_stretch_audio_no_pitch_correction,
+                                                      label="no_pitch_correction", name="no_pitch_correction")
+      set_tooltip(self.checkbox_no_pitch_correction, "Skips pitch correction step when stretching audio. " + \
+                                                     "Requires --stretch_audio to be set, otherwise " + \
+                                                     "does nothing.")
+      
+      self.button_save = wx.Button(self, label="Save")
+      self.button_save.Bind(wx.EVT_BUTTON, self.save_settings)
+      self.button_cancel = wx.Button(self, label="Cancel")
+      self.button_cancel.Bind(wx.EVT_BUTTON, lambda event: self.EndModal(0))
+      
+      sizer_dialog = wx.BoxSizer(wx.VERTICAL)
+      sizer_output_dir = wx.BoxSizer(wx.HORIZONTAL)
+      sizer_alignment_dir = wx.BoxSizer(wx.HORIZONTAL)
+      sizer_prepend = wx.BoxSizer(wx.HORIZONTAL)
+      sizer_stretch_audio_no_pitch_correction_outer = wx.BoxSizer(wx.HORIZONTAL)
+      sizer_stretch_audio_no_pitch_correction_inner = wx.BoxSizer(wx.VERTICAL)
+      sizer_save_cancel = wx.BoxSizer(wx.HORIZONTAL)
+      
+      # Configure layout with nested Box Sizers:
+      #
+      # Frame
+      #   sizer_dialog
+      #     text_header
+      #     sizer_output_dir
+      #       static_box_sizer_output
+      #         dir_picker_output
+      #     sizer_alignment_dir
+      #       static_box_sizer_alignment
+      #         dir_picker_alignment
+      #     sizer_prepend
+      #       text_prepend
+      #       text_ctrl_prepend
+      #     sizer_stretch_audio_no_pitch_correction_outer
+      #       panel_stretch_audio_no_pitch_correction
+      #         sizer_stretch_audio_no_pitch_correction_inner
+      #           checkbox_stretch_audio
+      #           checkbox_no_pitch_correction
+      #     sizer_save_cancel
+      #       button_save
+      #       button_cancel
+      #
+      self.SetSizer(sizer_dialog)
+      sizer_dialog.Add(self.text_header, 0, wx.ALL, 5)
+      sizer_dialog.Add(sizer_output_dir, 1, wx.LEFT|wx.RIGHT|wx.EXPAND, 2)
+      sizer_dialog.Add(sizer_alignment_dir, 1, wx.LEFT|wx.RIGHT|wx.EXPAND, 2)
+      sizer_dialog.Add(sizer_prepend, 1, wx.LEFT|wx.EXPAND, 5)
+      sizer_dialog.Add(sizer_stretch_audio_no_pitch_correction_outer, 1, wx.LEFT|wx.EXPAND, 5)
+      sizer_stretch_audio_no_pitch_correction_outer.Add(panel_stretch_audio_no_pitch_correction,
+                                                        1, wx.LEFT|wx.EXPAND, 5)
+      sizer_stretch_audio_no_pitch_correction_outer.Add((0, 0), 2, wx.EXPAND, 5)  # spacer
+      sizer_dialog.Add(sizer_save_cancel, 2, wx.BOTTOM|wx.EXPAND, 5)
+      sizer_prepend.Add(self.text_prepend, 0, wx.ALL|wx.ALIGN_CENTER_VERTICAL, 5)
+      sizer_prepend.Add(self.text_ctrl_prepend, 0, wx.ALIGN_CENTER_VERTICAL, 5)
+      sizer_output_dir.Add(self.static_box_sizer_output, 1, wx.LEFT|wx.RIGHT|wx.ALIGN_CENTER_VERTICAL, 5)
+      self.static_box_sizer_output.Add(self.dir_picker_output, 1, wx.EXPAND)
+      sizer_alignment_dir.Add(self.static_box_sizer_alignment, 1, wx.LEFT|wx.RIGHT|wx.ALIGN_CENTER_VERTICAL, 5)
+      self.static_box_sizer_alignment.Add(self.dir_picker_alignment, 1, wx.EXPAND)
+      panel_stretch_audio_no_pitch_correction.SetSizer(sizer_stretch_audio_no_pitch_correction_inner)
+      sizer_stretch_audio_no_pitch_correction_inner.Add(self.checkbox_stretch_audio, 0, wx.ALL, 5)
+      sizer_stretch_audio_no_pitch_correction_inner.Add(self.checkbox_no_pitch_correction, 0, wx.ALL, 5)
+      sizer_save_cancel.Add((0, 0), 3, wx.EXPAND, 5)  # spacer
+      sizer_save_cancel.Add(self.button_save, 0, wx.ALL|wx.ALIGN_CENTER_VERTICAL, 5)
+      sizer_save_cancel.Add((0, 0), 2, wx.EXPAND, 5)  # spacer
+      sizer_save_cancel.Add(self.button_cancel, 0, wx.ALL|wx.ALIGN_CENTER_VERTICAL, 5)
+      sizer_save_cancel.Add((0, 0), 3, wx.EXPAND, 5)  # spacer
+      
+      # centers GUI on the screen
+      self.Centre(wx.BOTH)
+      
+      # cache dictionaries mapping setting names to widget setter and getter functions
+      self.setting_getters = {}
+      self.setting_setters = {}
+      for child in itertools.chain(self.GetChildren(),
+                                   panel_stretch_audio_no_pitch_correction.GetChildren()):
+        child_class_name = child.GetClassName()
+        child_name = child.GetName()
+        if child_class_name == "wxDirPickerCtrl":
+          self.setting_getters[child_name] = child.GetPath
+          self.setting_setters[child_name] = child.SetPath
+        if child_class_name in ["wxCheckBox"]:
+          self.setting_getters[child_name] = child.GetValue
+          self.setting_setters[child_name] = child.SetValue
+        if child_class_name in ["wxTextCtrl"]:
+          self.setting_getters[child_name] = child.GetValue
+          self.setting_setters[child_name] = lambda value, child=child: child.SetValue(str(value))
+      self.setting_names = self.setting_getters.keys()
+      
+      # initialize setting widgets to saved config values
+      self.config_path = config_path
+      config_file_settings = read_config_file(self.config_path)
+      for setting_name in self.setting_names:
+        self.setting_setters[setting_name](config_file_settings[setting_name])
 
-def settings_gui(config_path: Path):
-  settings = read_config_file(config_path)
-  layout = [[sg.Text('Check tooltips (i.e. mouse-over text) for descriptions:')],
-            [sg.Column([[sg.Text('extension:', size=(10, 1.2), pad=(1,5)),
-                         sg.Input(default_text=str(settings['extension']), size=(8, 1.2), pad=(10,5), key='extension',
-                                  tooltip='File type of output video (e.g. mkv). When set to "copy", copies the ' + \
-                                          'file type of the corresponding input video. Default is "copy".')]])],
-            [sg.Column([[sg.Text('prepend:', size=(8, 1.2), pad=(1,5)),
-                         sg.Input(default_text=str(settings['prepend']), size=(8, 1.2), pad=(10,5), key='prepend',
-                                  tooltip='Output file name prepend text. Default is "ad_"')]])],
-            [sg.Column([[sg.Text('output_dir:', size=(10, 1.2), pad=(1,5)),
-                         sg.Input(default_text=str(settings['output_dir']), size=(22, 1.2), pad=(10,5), key='output_dir',
-                                  tooltip='Directory combined output media is saved to. Default is "videos_with_ad"'),
-                                  sg.FolderBrowse(button_text="Browse Folder", key='output_browse')]])],
-            [sg.Column([[sg.Text('alignment_dir:', size=(13, 1.2), pad=(1,5)),
-                         sg.Input(default_text=str(settings['alignment_dir']), size=(22, 1.2), pad=(10,5), key='alignment_dir',
-                                  tooltip='Directory alignment data and plots are saved to. Default is "alignment_plots"'),
-                         sg.FolderBrowse(button_text="Browse Folder", key='alignment_browse')]], pad=(2,7))],
-            [sg.Column([[sg.Text('smoothness:', size=(12, 1), pad=(1,5)),
-                         sg.Input(default_text=str(settings['smoothness']), size=(8, 1.2), pad=(10,5), key='smoothness',
-                                  tooltip='Lower values make the alignment more accurate when there are skips ' + \
-                                          '(e.g. describer pauses), but also make it more likely to misalign. ' + \
-                                          'Default is 50.')]])],
-            [sg.Checkbox('stretch_audio', default=settings['stretch_audio'], key='stretch_audio', change_submits=True,
-                         tooltip='Stretches the input audio to fit the input video. ' + \
-                                 'Default is to stretch the video to fit the audio.')],
-            [sg.Checkbox('keep_non_ad', default=settings['keep_non_ad'], key='keep_non_ad',
-                         disabled=not settings['stretch_audio'],
-                         tooltip='Tries to only replace segments with audio description. Useful if ' + \
-                                 'video\'s audio quality is better. Default is to replace all aligned audio. ' + \
-                                 'Requires --stretch_audio to be set, otherwise does nothing.')],
-            [sg.Column([[sg.Text('boost:', size=(6, 1), pad=(1,5)),
-                         sg.Input(default_text=str(settings['boost']), size=(8, 1.2), pad=(10,5),
-                                  key='boost', disabled=not settings['stretch_audio'],
-                                  tooltip='Boost (or quieten) description volume. Units are decibels (dB), so ' + \
-                                          '-3 makes the describer about 2x quieter, while 3 makes them 2x louder. ' + \
-                                          'Requires --stretch_audio to be set, otherwise does nothing.')]])],
-            [sg.Column([[sg.Text('ad_detect_sensitivity:', size=(21, 1.2), pad=(2,5)),
-                         sg.Input(default_text=str(settings['ad_detect_sensitivity']), size=(8, 1.2), pad=(10,5),
-                                  key='ad_detect_sensitivity', disabled=not settings['stretch_audio'],
-                                  tooltip='Audio description detection sensitivity ratio. Higher values make ' + \
-                                          '--keep_non_ad more likely to replace aligned audio. Default is 0.6')]])],
-            [sg.Column([[sg.Text('boost_sensitivity:', size=(17, 1.2), pad=(1,5)),
-                         sg.Input(default_text=str(settings['boost_sensitivity']), size=(8, 1.2), pad=(10,5),
-                                  key='boost_sensitivity', disabled=not settings['stretch_audio'],
-                                  tooltip='Higher values make --boost less likely to miss a description, but ' + \
-                                          'also make it more likely to boost non-description audio. Default is 0.4')]])],
-            [sg.Checkbox('no_pitch_correction', default=settings['no_pitch_correction'], key='no_pitch_correction',
-                         disabled=not settings['stretch_audio'],
-                         tooltip='Skips pitch correction step when stretching audio. ' + \
-                                 'Requires --stretch_audio to be set, otherwise does nothing.')],
-            [sg.Column([[sg.Submit('Save', pad=(40,3)),
-                         sg.Button('Cancel')]], pad=((135,3),10))]]
-  settings_window = sg.Window('Settings - describealign', layout, font=('Arial', 16), finalize=True)
-  settings_window['extension'].set_focus()
-  while True:
-    event, values = settings_window.read()
-    if event in (sg.WIN_CLOSED, 'Cancel') or settings_window.TKrootDestroyed:
-      break
-    if event == 'stretch_audio':
-      # work around bug in PySimpleGUIWx's InputText Update function where enabling/disabling are flipped
-      if IS_RUNNING_WINDOWS:
-        settings_window['boost'].Update(disabled = values['stretch_audio'])
-        settings_window['ad_detect_sensitivity'].Update(disabled = values['stretch_audio'])
-        settings_window['boost_sensitivity'].Update(disabled = values['stretch_audio'])
+      # initialize stretch_audio subsettings to be disabled/enabled
+      self.update_stretch_audio_subsettings()
+      
+      set_background_color(self, is_dark)
+      if sum(self.checkbox_stretch_audio.GetForegroundColour()[:3]) < 350:
+        panel_stretch_audio_no_pitch_correction.SetBackgroundColour(gui_background_color_light)
+    
+    def update_stretch_audio_subsettings(self, event=None):
+      subsettings = [self.checkbox_no_pitch_correction]
+      if self.checkbox_stretch_audio.IsChecked():
+        for subsetting in subsettings:
+          subsetting.Enable()
       else:
-        settings_window['boost'].Update(disabled = not values['stretch_audio'])
-        settings_window['ad_detect_sensitivity'].Update(disabled = not values['stretch_audio'])
-        settings_window['boost_sensitivity'].Update(disabled = not values['stretch_audio'])
-      settings_window['keep_non_ad'].Update(disabled = not values['stretch_audio'])
-      settings_window['no_pitch_correction'].Update(disabled = not values['stretch_audio'])
-    if event == 'Save':
-      settings = values.copy()
-      del settings['output_browse']
-      del settings['alignment_browse']
-      write_config_file(config_path, settings)
-      break
-  settings_window.close()
+        for subsetting in subsettings:
+          subsetting.Disable()
+    
+    def save_settings(self, event):
+      settings = {}
+      for setting_name in self.setting_names:
+        settings[setting_name] = self.setting_getters[setting_name]()
+      write_config_file(self.config_path, settings)
+      self.EndModal(0)
+
+  class QueueWriter(io.TextIOWrapper):
+    def __init__(self, queue) -> None:
+      super().__init__(buffer=io.BytesIO())
+      self._queue = queue
+      
+    def write(self, s: str) -> int:
+      self._queue.put(s)
+      return len(s)
 
-class QueueWriter(io.TextIOWrapper):
-  def __init__(self, queue) -> None:
-    super().__init__(buffer=io.BytesIO())
-    self._queue = queue
+  def combine_print_exceptions(print_queue, *args, **kwargs):
+    writer = QueueWriter(print_queue)
+    with redirect_stdout(writer), redirect_stderr(writer):
+      try:
+        combine(*args, **kwargs)
+      except Exception:
+        print("  ERROR: exception raised")
+        traceback.print_exc()
+
+  class FrameCombine(wx.Frame):
+    def __init__(self, parent, config_path, video_files, audio_files, is_dark):
+      wx.Frame.__init__(self, parent, title="Combining - describealign", size=wx.Size(800,600))
+      # setting the GUI frame's font causes all contained elements to inherit that font by default
+      self.SetFont(wx.Font(*gui_font))
+      self.SetBackgroundColour(gui_background_color_dark if is_dark else gui_background_color_light)
+      # wrap all widgets within a panel to enable tab traversal (i.e. pressing tab to swap GUI focus)
+      self.panel0 = wx.Panel(self, style=wx.TAB_TRAVERSAL)
+      
+      self.text_ctrl_output = wx.TextCtrl(self.panel0, style=wx.TE_MULTILINE|wx.TE_READONLY|wx.TE_RICH)
+      
+      self.button_close = wx.Button(self.panel0, label="Close")
+      self.button_close.Bind(wx.EVT_BUTTON, self.attempt_close)
+      # also capture other close events such as alt+f4 or clicking the X in the top corner of the frame
+      self.Bind(wx.EVT_CLOSE, self.attempt_close)
+      
+      self.update_timer = wx.Timer(self)
+      self.Bind(wx.EVT_TIMER, self.update_gui, self.update_timer)
+      
+      sizer_panel_outer = wx.BoxSizer(wx.VERTICAL)
+      sizer_panel_inner = wx.BoxSizer(wx.VERTICAL)
+      sizer_close_button = wx.BoxSizer(wx.HORIZONTAL)
+      
+      # Configure layout with nested Box Sizers:
+      #
+      # Frame
+      #   sizer_panel_outer
+      #     panel0
+      #       sizer_panel_inner
+      #         text_ctrl_output
+      #         sizer_close_button
+      #           button_close
+      #
+      self.SetSizer(sizer_panel_outer)
+      sizer_panel_outer.Add(self.panel0, 1, wx.EXPAND|wx.ALL, 5)
+      self.panel0.SetSizer(sizer_panel_inner)
+      sizer_panel_inner.Add(self.text_ctrl_output, 1, wx.ALL|wx.EXPAND, 5)
+      sizer_panel_inner.Add(sizer_close_button, 0, wx.EXPAND, 5)
+      sizer_close_button.Add((0, 0), 1, wx.EXPAND, 5)  # spacer
+      sizer_close_button.Add(self.button_close, 0, wx.ALL, 5)
+      sizer_close_button.Add((0, 0), 1, wx.EXPAND, 5)  # spacer
+      
+      # centers GUI on the screen
+      self.Centre(wx.BOTH)
+      
+      set_background_color(self, is_dark)
+      
+      self.config_path = config_path
+      self.overwrite_last_line = False
+      self.display_line('Combining media files:')
+      self.text_ctrl_output.SetInsertionPoint(0)
+      
+      # launch combiner using settings from config file, redirecting its output to a queue
+      self.print_queue = multiprocessing.Queue()
+      settings = read_config_file(self.config_path)
+      settings.update({'yes':True})
+      self.combine_process = multiprocessing.Process(target=combine_print_exceptions,
+                                                     args=(self.print_queue, video_files, audio_files),
+                                                     kwargs=settings, daemon=True)
+      self.combine_process.start()
+      self.update_gui()
+    
+    def attempt_close(self, event):
+      if self.combine_process.is_alive():
+        dialog = wx.MessageDialog(self, "Warning: combiner is still running, stop it and close anyway?",
+                                  "Warning", wx.YES_NO|wx.ICON_WARNING)
+        response = dialog.ShowModal()
+        if (response == wx.ID_YES):
+          self.combine_process.terminate()
+          self.Destroy()
+        elif (response == wx.ID_NO):
+          # If the EVT_CLOSE came from the OS, let the OS know it didn't succeed
+          if event.GetEventType() == wx.EVT_CLOSE.evtType[0]:
+            event.Veto(True)
+      else:
+        self.Destroy()
+    
+    def set_last_line_color(self, color, line_start):
+      num_lines = self.text_ctrl_output.GetNumberOfLines()
+      end = self.text_ctrl_output.GetLastPosition()
+      self.text_ctrl_output.SetStyle(line_start, end, wx.TextAttr("black", color))
     
-  def write(self, s: str) -> int:
-    self._queue.put(s)
-    return len(s)
+    def display_line(self, line):
+      if self.overwrite_last_line:
+        # skip the empty line following lines ending in "\r"
+        if line == "":
+          return
+        num_lines = self.text_ctrl_output.GetNumberOfLines()
+        start = self.text_ctrl_output.XYToPosition(0,num_lines-2)
+        end = self.text_ctrl_output.GetLastPosition()
+        self.text_ctrl_output.Remove(start, end)
+        self.overwrite_last_line = False
+      if line[-1:] == "\r":
+        self.overwrite_last_line = True
+        line = line[:-1].rstrip(' ') + "\r"
+      line_start = self.text_ctrl_output.GetLastPosition()
+      self.text_ctrl_output.AppendText(line)
+      # highlight warnings by changing their background color to light orange
+      if line[:10] == "  WARNING:":
+        self.set_last_line_color(wx.Colour(255, 188, 64), line_start)
+      # highlight errors by changing their background color to red
+      if line[:8] == "  ERROR:":
+        self.set_last_line_color(wx.Colour(255, 128, 128), line_start)
+    
+    def update_gui(self, event=None):
+      lines = []
+      while not self.print_queue.empty():
+        lines.append(self.print_queue.get())
+      if len(lines) > 0:
+        cursor_position = self.text_ctrl_output.GetInsertionPoint()
+        self.text_ctrl_output.Freeze()
+        for line in lines:
+          self.display_line(line)
+        self.text_ctrl_output.SetInsertionPoint(cursor_position)
+        self.text_ctrl_output.Thaw()
+      self.update_timer.StartOnce(gui_update_interval_ms)
 
-def combine_print_exceptions(print_queue, *args, **kwargs):
-  writer = QueueWriter(print_queue)
-  with redirect_stdout(writer), redirect_stderr(writer):
+  def migrate_config(old_path: Optional[Path], new_path: Path) -> None:
+    """
+    Migrate configuration from old location.
+    
+    Only runs if the old_path exists but new_path does not
+    """
+    if new_path.exists() or not old_path or not old_path.exists():
+      return
+    
+    old_data = old_path.read_text(encoding='utf-8')
+    new_path.write_text(old_data, encoding='utf-8')
+    print(f"Configuration migrated to {new_path}")
     try:
-      combine(*args, **kwargs)
-    except Exception:
-      traceback.print_exc()
+      old_path.unlink()
+    except OSError as exc:
+      print("Failed to remove old config:", *traceback.format_exception_only(exc))
+    else:
+      print("Successfully removed old config file.")
 
-def combine_gui(video_files, audio_files, config_path):
-  output_textbox = sg.Multiline(size=(80,30), key='-OUTPUT-')
-  layout = [[output_textbox],
-            [sg.Button('Close', pad=(360,5))]]
-  combine_window = sg.Window('Combining - describealign', layout, font=('Arial', 16),
-                             disable_close=True, finalize=True)
-  output_textbox.update('Combining media files:', append=True)
-  print_queue = multiprocessing.Queue()
-  
-  settings = read_config_file(config_path)
-  settings.update({'yes':True})
-  proc = multiprocessing.Process(target=combine_print_exceptions,
-                                 args=(print_queue, video_files, audio_files),
-                                 kwargs=settings, daemon=True)
-  proc.start()
-  while True:
-    # if the script isn't running anymore, re-enable the default close window button
-    if not proc.is_alive():
-      combine_window.DisableClose = False
-    if not print_queue.empty():
-      if IS_RUNNING_WINDOWS:
-        cursor_position = output_textbox.WxTextCtrl.GetInsertionPoint()
-      output_textbox.update(print_queue.get(), append=True)
-      if IS_RUNNING_WINDOWS:
-        output_textbox.WxTextCtrl.SetInsertionPoint(cursor_position)
-    event, values = combine_window.read(timeout=100)
-    # window closed event isn't always emitted, so also manually check window status
-    if event == sg.WIN_CLOSED or combine_window.TKrootDestroyed:
-      if proc.is_alive():
-        proc.terminate()
-      break
-    if event == 'Close':
-      if not proc.is_alive():
-        combine_window.DisableClose = False
-        break
-      selection = sg.PopupYesNo('Combiner is still running, stop it and close anyway?')
-      if selection != 'Yes':
+  class ListCtrlDropTarget(wx.FileDropTarget):
+    def __init__(self, list_ctrl, parent_frame):
+      super().__init__()
+      self.list_ctrl = list_ctrl
+      self.parent_frame = parent_frame
+    
+    def expand_folders(self, files):
+      expanded_files = []
+      for file in files:
+        if os.path.isdir(file):
+          for dir, subdirs, dir_files in os.walk(file):
+            for dir_file in dir_files:
+              expanded_files.append(os.path.join(dir, dir_file))
+        else:
+          expanded_files.append(file)
+      return expanded_files
+    
+    def OnDropFiles(self, x, y, files):
+      files = self.expand_folders(files)
+      valid_file_types = self.parent_frame.list_ctrl_file_types_drop[self.list_ctrl]
+      files = [file for file in files if os.path.splitext(file)[-1][1:] in valid_file_types]
+      self.parent_frame.populate_list_ctrl(self.list_ctrl, natsort.os_sorted(files))
+      return True
+
+  def get_children(window):
+    children = list(window.GetChildren())
+    subchildren = [subchild for child in children for subchild in get_children(child)]
+    return children + subchildren
+
+  def set_background_color(window, is_dark):
+    children = get_children(window)
+    for window in children + [window]:
+      # modifying a CheckBox converts it into a Button, which would mess with screen readers
+      if isinstance(window, wx.CheckBox):
         continue
-      proc.terminate()
-      combine_window.DisableClose = False
-      break
-  combine_window.close()
+      if is_dark:
+        if isinstance(window, (wx.ListCtrl, wx.TextCtrl)):
+          window.SetBackgroundColour("Black")
+        elif isinstance(window, wx.Button):
+          window.SetBackgroundColour(tuple(x // 2 for x in gui_background_color_dark))          
+        else:
+          window.SetBackgroundColour(gui_background_color_dark)
+      window.SetForegroundColour("White" if is_dark else "Black")
 
-def migrate_config(old_path: Optional[Path], new_path: Path) -> None:
-  """
-  Migrate configuration from old location.
-  
-  Only runs if the old_path exists but new_path does not
-  """
-  if new_path.exists() or not old_path or not old_path.exists():
-    return
-  
-  old_data = old_path.read_text(encoding='utf-8')
-  new_path.write_text(old_data, encoding='utf-8')
-  print(f"Configuration migrated to {new_path}")
-  try:
-    old_path.unlink()
-  except OSError as exc:
-    print("Failed to remove old config:", *traceback.format_exception_only(exc))
-  else:
-    print("Successfully removed old config file.")
+  class FrameMain(wx.Frame):
+    def __init__(self, parent):
+      wx.Frame.__init__(self, parent, title=f"describealign v{__version__}", size=wx.Size(800, 500))
+      # setting the GUI frame's font causes all contained elements to inherit that font by default
+      self.SetFont(wx.Font(*gui_font))
+      appearance = wx.SystemSettings.GetAppearance()
+      self.is_dark = appearance.IsDark() or appearance.IsUsingDarkBackground()
+      self.SetBackgroundColour(gui_background_color_dark if self.is_dark else gui_background_color_light)
+      
+      # wrap all widgets within a panel to enable tab traversal (i.e. pressing tab to swap GUI focus)
+      self.panel0 = wx.Panel(self, style=wx.TAB_TRAVERSAL)
+      
+      self.text_header = wx.StaticText(self.panel0, label="Select media files to combine:")
+      self.text_header.SetFont(self.text_header.GetFont().Scale(1.7))
+      
+      # Video Input selection and display row of GUI
+      self.static_box_sizer_video = wx.StaticBoxSizer(wx.HORIZONTAL, self.panel0, "Video Input")
+      self.list_ctrl_video = self.init_list_ctrl(self.static_box_sizer_video.GetStaticBox(),
+                                                 "Drag and Drop Videos Here or Press Browse Video")
+      set_tooltip(self.list_ctrl_video, "Video filenames are listed here in the sorted order they will " + \
+                                        "be used as input. Drag and Drop or press Browse to overwrite.")
+      self.button_browse_video = wx.Button(self.static_box_sizer_video.GetStaticBox(), label="Browse Video")
+      set_tooltip(self.button_browse_video, "Select one or more video files as input.")
+      self.button_browse_video.Bind(wx.EVT_BUTTON, lambda event: self.browse_files(self.list_ctrl_video))
+      
+      # Audio Input selection and display row of GUI
+      self.static_box_sizer_audio = wx.StaticBoxSizer(wx.HORIZONTAL, self.panel0, "Audio Input")
+      self.list_ctrl_audio = self.init_list_ctrl(self.static_box_sizer_audio.GetStaticBox(),
+                                                 "Drag and Drop Audio Here or Press Browse Audio")
+      set_tooltip(self.list_ctrl_audio, "Audio filenames are listed here in the sorted order they will " + \
+                                        "be used as input. Drag and Drop or press Browse to overwrite.")
+      self.button_browse_audio = wx.Button(self.static_box_sizer_audio.GetStaticBox(), label="Browse Audio")
+      set_tooltip(self.button_browse_audio, "Select one or more audio files as input.")
+      self.button_browse_audio.Bind(wx.EVT_BUTTON, lambda event: self.browse_files(self.list_ctrl_audio))
+      
+      self.button_combine = wx.Button(self.panel0, label="Combine")
+      set_tooltip(self.button_combine, "Combine selected video and audio files.")
+      self.button_combine.Bind(wx.EVT_BUTTON, self.open_combine)
+      self.button_settings = wx.Button(self.panel0, label="Settings")
+      set_tooltip(self.button_settings, "Edit settings for the GUI and algorithm.")
+      self.button_settings.Bind(wx.EVT_BUTTON, self.open_settings)
+      
+      sizer_panel_outer = wx.BoxSizer(wx.VERTICAL)
+      sizer_panel_inner = wx.BoxSizer(wx.VERTICAL)
+      sizer_header = wx.BoxSizer(wx.HORIZONTAL)
+      sizer_video = wx.BoxSizer(wx.HORIZONTAL)
+      sizer_audio = wx.BoxSizer(wx.HORIZONTAL)
+      sizer_combine_settings = wx.BoxSizer(wx.HORIZONTAL)
+      
+      # Configure layout with nested Box Sizers:
+      #
+      # Frame
+      #   sizer_panel_outer
+      #     panel0
+      #       sizer_panel_inner
+      #         sizer_header
+      #           text_header
+      #         sizer_video
+      #           list_ctrl_video
+      #           button_browse_video
+      #         sizer_audio
+      #           list_ctrl_audio
+      #           button_browse_audio
+      #         sizer_combine_settings
+      #           button_combine
+      #           button_settings
+      #
+      self.SetSizer(sizer_panel_outer)
+      sizer_panel_outer.Add(self.panel0, 1, wx.EXPAND|wx.ALL, 5)
+      self.panel0.SetSizer(sizer_panel_inner)
+      sizer_panel_inner.Add(sizer_header, 3, wx.EXPAND, 5)
+      sizer_panel_inner.Add(sizer_video, 9, wx.EXPAND, 5)
+      sizer_panel_inner.Add(sizer_audio, 9, wx.TOP|wx.EXPAND, 3)
+      sizer_panel_inner.Add(sizer_combine_settings, 3, wx.EXPAND, 5)
+      sizer_header.Add(self.text_header, 0, wx.ALL|wx.ALIGN_CENTER_VERTICAL, 5)
+      sizer_video.Add(self.static_box_sizer_video, 1, wx.LEFT|wx.RIGHT|wx.EXPAND, 3)
+      self.static_box_sizer_video.Add(self.list_ctrl_video, 1, wx.BOTTOM|wx.EXPAND, 2)
+      self.static_box_sizer_video.Add(self.button_browse_video, 0,
+                                      wx.LEFT|wx.BOTTOM|wx.RIGHT|wx.ALIGN_CENTER_VERTICAL, 10)
+      sizer_audio.Add(self.static_box_sizer_audio, 1, wx.LEFT|wx.RIGHT|wx.EXPAND, 3)
+      self.static_box_sizer_audio.Add(self.list_ctrl_audio, 1, wx.BOTTOM|wx.EXPAND, 2)
+      self.static_box_sizer_audio.Add(self.button_browse_audio, 0,
+                                      wx.LEFT|wx.BOTTOM|wx.RIGHT|wx.ALIGN_CENTER_VERTICAL, 10)
+      sizer_combine_settings.Add((0, 0), 7, wx.EXPAND, 5)  # spacer
+      sizer_combine_settings.Add(self.button_combine, 0, wx.ALL|wx.ALIGN_CENTER_VERTICAL, 5)
+      sizer_combine_settings.Add((0, 0), 2, wx.EXPAND, 5)  # spacer
+      sizer_combine_settings.Add(self.button_settings, 0, wx.ALL|wx.ALIGN_CENTER_VERTICAL, 5)
+      sizer_combine_settings.Add((0, 0), 7, wx.EXPAND, 5)  # spacer
+      
+      # centers GUI on the screen
+      self.Centre(wx.BOTH)
+      
+      all_video_file_types = [('All Video File Types', '*.' + ';*.'.join(VIDEO_EXTENSIONS)),]
+      all_audio_file_types = [('All Audio File Types', '*.' + ';*.'.join(AUDIO_EXTENSIONS)),]
+      all_video_and_audio_file_types = [('All Video and Audio File Types',
+                                         '*.' + ';*.'.join(VIDEO_EXTENSIONS | AUDIO_EXTENSIONS)),]
+      self.video_file_types = [(ext, f"*.{ext}") for ext in VIDEO_EXTENSIONS]
+      self.audio_file_types = [(ext, f"*.{ext}") for ext in AUDIO_EXTENSIONS]
+      self.video_and_audio_file_types = self.video_file_types + self.audio_file_types
+      self.video_file_types = all_video_file_types + self.video_file_types
+      self.audio_file_types = all_audio_file_types + self.audio_file_types
+      self.video_and_audio_file_types = all_video_file_types + all_video_and_audio_file_types + \
+                                        self.video_and_audio_file_types
+      self.video_file_types = '|'.join([f'{type[0]} ({type[1]})|{type[1]}' for type in self.video_file_types])
+      self.audio_file_types = '|'.join([f'{type[0]} ({type[1]})|{type[1]}' for type in self.audio_file_types])
+      self.video_and_audio_file_types = '|'.join([f'{type[0]} ({type[1]})|{type[1]}' for type \
+                                                  in self.video_and_audio_file_types])
+      
+      # track the allowed file types and selected files' full paths for each List Ctrl
+      self.list_ctrl_file_types_browse = {self.list_ctrl_video: self.video_and_audio_file_types,
+                                          self.list_ctrl_audio: self.audio_file_types}
+      self.list_ctrl_file_types_drop = {self.list_ctrl_video: self.video_file_types,
+                                        self.list_ctrl_audio: self.audio_file_types}
+      self.list_ctrl_files_selected = {self.list_ctrl_video: [],
+                                       self.list_ctrl_audio: []}
+      
+      self.config_path = self.get_config()
+      
+      set_background_color(self, self.is_dark)
+    
+    def init_list_ctrl(self, parent_panel, default_text):
+      list_ctrl = wx.ListCtrl(parent_panel, style=wx.LC_NO_HEADER|wx.LC_REPORT|wx.BORDER_SUNKEN|wx.HSCROLL)
+      list_ctrl.EnableSystemTheme(False)  # get rid of vertical grid lines on Windows
+      list_ctrl.SetMinSize(wx.Size(-1,80))
+      list_ctrl.SetDropTarget(ListCtrlDropTarget(list_ctrl, self))
+      list_ctrl.InsertColumn(0, "")
+      list_ctrl.InsertItem(0, default_text)
+      list_ctrl.SetColumnWidth(0, wx.LIST_AUTOSIZE)
+      list_ctrl.Bind(wx.EVT_CHAR, self.delete_from_list_ctrl)
+      return list_ctrl
+    
+    def populate_list_ctrl(self, list_ctrl, files):
+      self.list_ctrl_files_selected[list_ctrl] = files
+      if len(files) == 0:
+        files = ["No files with valid file types found"]
+      list_ctrl.DeleteAllItems()
+      list_ctrl.DeleteAllColumns()
+      list_ctrl.InsertColumn(0, "")
+      for i, file in enumerate(files):
+        list_ctrl.InsertItem(i, os.path.basename(file))
+      list_ctrl.SetColumnWidth(0, wx.LIST_AUTOSIZE)
+    
+    def browse_files(self, list_ctrl):
+      dialog = wx.FileDialog(self, wildcard=self.list_ctrl_file_types_browse[list_ctrl], style=wx.FD_MULTIPLE)
+      if dialog.ShowModal() == wx.ID_OK:
+        files = dialog.GetPaths()
+        self.populate_list_ctrl(list_ctrl, files)
+    
+    def delete_from_list_ctrl(self, event):
+      if event.GetKeyCode() == wx.WXK_DELETE:
+        list_ctrl = event.GetEventObject()
+        item_index = list_ctrl.GetFirstSelected()
+        if item_index == -1:
+          item_index = list_ctrl.GetFocusedItem()
+        items_to_delete = []
+        while item_index != -1:
+          items_to_delete.append(item_index)
+          item_index = list_ctrl.GetNextSelected(item_index)
+        for item_index in items_to_delete[::-1]:
+          if len(self.list_ctrl_files_selected[list_ctrl]) != 0:
+            list_ctrl.DeleteItem(item_index)
+            del self.list_ctrl_files_selected[list_ctrl][item_index]
+      else:
+        event.Skip()
+    
+    def open_combine(self, event):
+      video_files = self.list_ctrl_files_selected[self.list_ctrl_video]
+      audio_files = self.list_ctrl_files_selected[self.list_ctrl_audio]
+      if len(video_files) == 0:
+        error_dialog = wx.MessageDialog(self, "Error: no video input selected.", "Error", wx.OK|wx.ICON_ERROR)
+        error_dialog.ShowModal()
+      elif len(audio_files) == 0:
+        error_dialog = wx.MessageDialog(self, "Error: no audio input selected.", "Error", wx.OK|wx.ICON_ERROR)
+        error_dialog.ShowModal()
+      elif len(video_files) != len(audio_files):
+        error_dialog = wx.MessageDialog(self, f"Error: different numbers of video ({len(video_files)}) " + \
+                                              f"and audio ({len(audio_files)}) inputs.",
+                                        "Error", wx.OK|wx.ICON_ERROR)
+        error_dialog.ShowModal()
+      else:
+        frame_combine = FrameCombine(None, self.config_path, video_files, audio_files, self.is_dark)
+        self.list_ctrl_video.SetFocus()
+        frame_combine.Show()
+    
+    def open_settings(self, event):
+      dialog_settings = DialogSettings(None, self.config_path, self.is_dark)
+      dialog_settings.ShowModal()
+      dialog_settings.Destroy()
+    
+    def get_config(self):
+      config_path = platformdirs.user_config_path(appname='describealign', appauthor=False,
+                                                  ensure_exists=True) / 'config.ini'
+      old_paths = [
+        # Place in chronological order (oldest -> newest)
+        Path(__file__).resolve().parent / 'config.ini',
+        platformdirs.user_config_path(appname='describealign', ensure_exists=True) / 'config.ini',
+      ]
+      # Get newest existent path
+      old_config = next((file for file in reversed(old_paths) if file.exists()), None,)
+      try:
+        migrate_config(old_config, config_path)
+      except OSError as exc:
+        print(f"Error migrating old config:", *traceback.format_exception_only(exc))
+        print(f"Old config left in place at {old_config}")
+      return config_path
 
-def main_gui():
-  config_path = platformdirs.user_config_path(appname='describealign', appauthor=False, ensure_exists=True) / 'config.ini'
-  old_paths = [
-    # Place in chronological order (oldest -> newest)
-    Path(__file__).resolve().parent / 'config.ini',
-    platformdirs.user_config_path(appname='describealign', ensure_exists=True) / 'config.ini',
-  ]
-  
-  # Get newest existent path
-  old_config = next(
-    (
-      file 
-      for file in reversed(old_paths)
-      if file.exists()
-    ), 
-    None,
-  )
-  
+def get_version_hash(filename):
   try:
-    migrate_config(old_config, config_path)
-  except OSError as exc:
-    print(f"Error migrating old config:", *traceback.format_exception_only(exc))
-    print(f"Old config left in place at {old_config}")
-  
-  sg.theme('Light Blue 2')
-  
-  filetype_sep = ';' if IS_RUNNING_WINDOWS else ' '
-  all_audio_file_types = [('All Audio File Types', '*.' + f'{filetype_sep}*.'.join(AUDIO_EXTENSIONS)),]
-  all_video_file_types = [('All Video File Types', '*.' + f'{filetype_sep}*.'.join(VIDEO_EXTENSIONS)),]
-  all_video_and_audio_file_types = [('All Video and Audio File Types',
-                                     '*.' + f'{filetype_sep}*.'.join(VIDEO_EXTENSIONS | AUDIO_EXTENSIONS)),]
-  audio_file_types = [(ext, f"*.{ext}") for ext in AUDIO_EXTENSIONS]
-  video_and_audio_file_types = [(ext, f"*.{ext}") for ext in VIDEO_EXTENSIONS] + audio_file_types
-  audio_file_types = all_audio_file_types + audio_file_types
-  video_and_audio_file_types = all_video_file_types + all_video_and_audio_file_types + video_and_audio_file_types
-  # work around bug in PySimpleGUIWx's convert_tkinter_filetypes_to_wx function
-  if IS_RUNNING_WINDOWS:
-    file_fix = lambda file_types: file_types[:1] + [(f'|{type[0]}', type[1]) for type in file_types[1:]]
-    audio_file_types = file_fix(audio_file_types)
-    video_and_audio_file_types = file_fix(video_and_audio_file_types)
-  
-  layout = [[sg.Text('Select media files to combine:', size=(40, 2), font=('Arial', 20), pad=(3,15))],
-            [sg.Column([[sg.Text('Video Input:', size=(11, 2), pad=(1,5)),
-                         sg.Input(size=(35, 1.2), pad=(10,5), key='-VIDEO_FILES-',
-                                  tooltip='List video filenames here, in order, separated by semicolons'),
-                         sg.FilesBrowse(button_text="Browse Video",
-                                        file_types=video_and_audio_file_types,
-                                        tooltip='Select one or more video files')]], pad=(2,7))],
-            [sg.Column([[sg.Text('Audio Input:', size=(11, 2), pad=(1,5)),
-                         sg.Input(size=(35, 1.2), pad=(10,5), key='-AUDIO_FILES-',
-                                  tooltip='List audio filenames here, in order, separated by semicolons'),
-                         sg.FilesBrowse(button_text="Browse Audio",
-                                        file_types=audio_file_types,
-                                        tooltip='Select one or more audio files')]], pad=(2,7))],
-            [sg.Column([[sg.Submit('Combine', pad=(40,3), tooltip='Combine selected video and audio files'),
-                         sg.Button('Settings', tooltip='Edit settings for the GUI and algorithm.')]],
-                         pad=((135,3),10))]]
-  window = sg.Window('describealign', layout, font=('Arial', 16), resizable=False, finalize=True)
-  window['-VIDEO_FILES-'].set_focus()
-  while True:
-    event, values = window.read()
-    if event == 'Combine':
-      if len(values['-VIDEO_FILES-']) == 0 or \
-         len(values['-AUDIO_FILES-']) == 0:
-        window.disable()
-        sg.Popup('Error: empty input field.', font=('Arial', 20))
-        window.enable()
-        window['-VIDEO_FILES-'].set_focus()
-        continue
-      video_files = values['-VIDEO_FILES-'].split(';')
-      if len(video_files) == 1:
-        video_files = video_files[0]
-      audio_files = values['-AUDIO_FILES-'].split(';')
-      if len(audio_files) == 1:
-        audio_files = audio_files[0]
-      window.disable()
-      combine_gui(video_files, audio_files, config_path)
-      window.enable()
-      window['-VIDEO_FILES-'].set_focus()
-    if event == 'Settings':
-      window.disable()
-      settings_gui(config_path)
-      window.enable()
-      window['-VIDEO_FILES-'].set_focus()
-    if event == sg.WIN_CLOSED:
-      break
-  window.close()
+    with open(filename, 'rb') as f:
+      data = f.read()
+      sha_hash = hashlib.sha1(data).hexdigest()
+    return sha_hash[:8]
+  except:
+    return "None"
 
 # Entry point for command line interaction, for example:
 # > describealign video.mp4 audio_desc.mp3
 def command_line_interface():
   if len(sys.argv) < 2:
-    # No args, run gui
-    print('No input arguments detected, starting GUI...')
-    main_gui()
-    sys.exit(0)
+    if wx is not None:
+      # No args, run gui
+      print('No input arguments detected, starting GUI...')
+      # the following line is necessary on MacOS X to fix the filectrlpicker
+      # https://docs.wxpython.org/wx.FileDialog.html#wx-filedialog
+      # https://github.com/wxWidgets/Phoenix/issues/2368
+      if platform.system() == 'Darwin':
+        wx.SystemOptions.SetOption('osx.openfiledialog.always-show-types', 1)
+      app = wx.App()
+      main_gui = FrameMain(None)
+      main_gui.Show()
+      app.MainLoop()
+      sys.exit(0)
+    else:
+      print("Can't launch GUI and arguments missing.\nGUI dependencies missing.")
   
-  parser = argparse.ArgumentParser(
-                          description="Replaces a video's sound with an audio description.",
-                          usage="describealign video_file.mp4 audio_file.mp3")
-  parser.add_argument("video", help='A video file or directory containing video files.', nargs='?', default=None)
-  parser.add_argument("audio", help='An audio file or directory containing audio files.', nargs='?', default=None)
-  parser.add_argument('--smoothness', type=float, default=50,
-                      help='Lower values make the alignment more accurate when there are skips ' + \
-                           '(e.g. describer pauses), but also make it more likely to misalign. ' + \
-                           'Default is 50.')
+  parser = argparse.ArgumentParser(description="Replaces a video's sound with an audio description.",
+                                   usage="describealign video_file.mp4 audio_file.mp3")
+  parser.add_argument("video", help='A video file or directory containing video files.',
+                      nargs='?', default=None)
+  parser.add_argument("audio", help='An audio file or directory containing audio files.',
+                      nargs='?', default=None)
   parser.add_argument('--stretch_audio', action='store_true',
                       help='Stretches the input audio to fit the input video. ' + \
-                           'Default is to stretch the video to fit the audio.')
-  parser.add_argument('--keep_non_ad', action='store_true',
-                      help='Tries to only replace segments with audio description. Useful if ' + \
-                           'video\'s audio quality is better. Default is to replace all aligned audio. ' + \
-                           'Requires --stretch_audio to be set, otherwise does nothing.')
-  parser.add_argument('--boost', type=float, default=0,
-                      help='Boost (or quieten) description volume. Units are decibels (dB), so ' + \
-                           '-3 makes the describer about 2x quieter, while 3 makes them 2x louder. ' + \
-                           'Requires --stretch_audio to be set, otherwise does nothing.')
-  parser.add_argument('--ad_detect_sensitivity', type=float, default=.6,
-                      help='Audio description detection sensitivity ratio. Higher values make ' + \
-                           '--keep_non_ad more likely to replace aligned audio. Default is 0.6')
-  parser.add_argument('--boost_sensitivity', type=float, default=.4,
-                      help='Higher values make --boost less likely to miss a description, but ' + \
-                           'also make it more likely to boost non-description audio. Default is 0.4')
+                           'Default is to stretch the video to fit the audio. ' + \
+                           'Keeps original video audio as secondary tracks. Slower ' + \
+                           'and uses more RAM when enabled, long videos may cause ' + \
+                           'paging or Out of Memory errors on low-RAM systems.')
   parser.add_argument('--yes', action='store_true',
                       help='Auto-skips user prompts asking to verify information.')
   parser.add_argument("--prepend", default="ad_", help='Output file name prepend text. Default is "ad_"')
@@ -1292,23 +1737,41 @@ def command_line_interface():
                       help='Directory combined output media is saved to. Default is "videos_with_ad"')
   parser.add_argument("--alignment_dir", default=default_alignment_dir,
                       help='Directory alignment data and plots are saved to. Default is "alignment_plots"')
-  parser.add_argument("--extension", default="copy",
-                      help='File type of output video (e.g. mkv). When set to "copy", copies the ' + \
-                           'file type of the corresponding input video. Default is "copy".')
   parser.add_argument("--install-ffmpeg", action="store_true",
-                      help="Install the required ffmpeg binaries and then exit. This is meant to be" + \
+                      help="Install the required ffmpeg binaries and then exit. This is meant to be " + \
                            "run from a privileged installer process (e.g. OS X Installer)")
+  parser.add_argument('--version', action='store_true',
+                      help='Checks and prints the installed version of describealign.')
   args = parser.parse_args()
   
-  if args.install_ffmpeg:
+  if args.version:
+    print(f"version: {__version__}")
+    if "__compiled__" in globals() or getattr(sys, 'frozen', False):
+      print("running from compiled binary")
+    else:
+      import importlib
+      cur_dir = os.getcwd()
+      if sys.path[0] == cur_dir:
+        # ignore describealign.py in current directory
+        del sys.path[0]
+        installed_spec = importlib.util.find_spec('describealign')
+        sys.path = [cur_dir] + sys.path
+      else:
+        installed_spec = importlib.util.find_spec('describealign')
+      this_script_path = os.path.abspath(__file__)
+      if installed_spec is None or (this_script_path != os.path.abspath(installed_spec.origin)):
+        print("running from downloaded .py file")
+      else:
+        print("running from installed package")
+      print(f"path: {this_script_path}")
+      print(f"content hash: {get_version_hash(this_script_path)}")
+  elif args.install_ffmpeg:
     # Make sure the file is world executable
     os.chmod(get_ffmpeg(), 0o755)
     os.chmod(get_ffprobe(), 0o755)
-  elif args.video or args.audio:
-    combine(args.video, args.audio, args.smoothness, args.stretch_audio, args.keep_non_ad,
-            args.boost, args.ad_detect_sensitivity, args.boost_sensitivity, args.yes,
-            args.prepend, args.no_pitch_correction, args.output_dir, args.alignment_dir,
-            args.extension)
+  elif args.video and args.audio:
+    combine(args.video, args.audio, args.stretch_audio, args.yes, args.prepend, 
+            args.no_pitch_correction, args.output_dir, args.alignment_dir)
   else:
     parser.print_usage()