modusa 0.4.20__py3-none-any.whl → 0.4.23__py3-none-any.whl

modusa/__init__.py

@@ -8,4 +8,4 @@ from modusa.tools import play, convert, record, save
 from modusa.tools import download
 from modusa.tools import load, load_ann
 
-__version__ = "0.4.20"
+__version__ = "0.4.23"

modusa/tools/ann_loader.py

@@ -6,7 +6,7 @@
 # Email: ankit0.anand0@gmail.com
 #---------------------------------
 
-def load_ann(path, clip=None):
+def load_ann(path, trim=None):
 	"""
 	Load annotation from audatity label
 	text file and also ctm file.
@@ -15,9 +15,9 @@ def load_ann(path, clip=None):
 	----------
 	path: str
 		- label text/ctm file path.
-	clip: tuple[number, number] | number | None
-		- Incase you clipped the audio signal, this parameter will help clip the annotation.
-		- If you clip the audio, say from (10, 20), set the clip to (10, 20).
+	trim: tuple[number, number] | number | None
+		- Incase you trimmed the audio signal, this parameter will help clip the annotation making sure that the timings are aligned to the trimmed audio.
+		- If you trimmed the audio, say from (10, 20), set the trim to (10, 20).
 		- Default: None
 
 	Returns
@@ -42,14 +42,14 @@ def load_ann(path, clip=None):
 	ann = [] # This will store the annotation
 	
 	# Clipping the annotation to match with the clipped audio
-	if clip is not None:
+	if trim is not None:
 		# Map clip input to the right format
-		if isinstance(clip, int or float):
-			clip = (0, clip)
-		elif isinstance(clip, tuple) and len(clip) > 1:
-			clip = (clip[0], clip[1])
+		if isinstance(trim, int or float):
+			trim = (0, trim)
+		elif isinstance(trim, tuple) and len(trim) > 1:
+			trim = (trim[0], trim[1])
 		else:
-			raise ValueError(f"Invalid clip type or length: {type(clip)}, len={len(clip)}")
+			raise ValueError(f"Invalid clip type or length: {type(trim)}, len={len(trim)}")
 	
 	if path.suffix == ".txt":
 		with open(str(path), "r") as f:
@@ -60,11 +60,11 @@ def load_ann(path, clip=None):
 				
 				# Incase user has clipped the audio signal, we adjust the annotation
 				# to match the clipped audio
-				if clip is not None:
-					offset = clip[0]
+				if trim is not None:
+					offset = trim[0]
 					# Clamp annotation to clip boundaries
-					new_start = max(start, clip[0]) - offset
-					new_end   = min(end, clip[1]) - offset
+					new_start = max(start, trim[0]) - offset
+					new_end   = min(end, trim[1]) - offset
 					
 					# only keep if there's still overlap
 					if new_start < new_end:
@@ -89,11 +89,11 @@ def load_ann(path, clip=None):
 				
 				# Incase user has clipped the audio signal, we adjust the annotation
 				# to match the clipped audio
-				if clip is not None:
-					offset = clip[0]
+				if trim is not None:
+					offset = trim[0]
 					# Clamp annotation to clip boundaries
-					new_start = max(start, clip[0]) - offset
-					new_end   = min(end, clip[1]) - offset
+					new_start = max(start, trim[0]) - offset
+					new_end   = min(end, trim[1]) - offset
 					
 					# only keep if there's still overlap
 					if new_start < new_end:

modusa/tools/audio_loader.py

@@ -1,106 +1,155 @@
 #!/usr/bin/env python3
 
 
-import soundfile as sf
-from scipy.signal import resample
+import subprocess
+import numpy as np
+import imageio_ffmpeg as ffmpeg
 from pathlib import Path
-import tempfile
-from scipy.signal import resample
-from .youtube_downloader import download
-from .audio_converter import convert
+import re
 
+def _get_audio_info_ffmpeg(path: Path):
+	"""
+	To get the original sampling rate and number of
+	channels of a given audio file by parsing the
+	metadata. (No extra tool required).
 
-def load(path, sr=None, clip=None):
+	Parameters
+	----------
+	audiofp: PathLike
+		- Audio filepath
+	
+	Returns
+	-------
+	int
+		- Original sampling rate (hz)
+	int
+		- Number of channels
 	"""
-	Loads audio file from various sources.
+	ffmpeg_exe = ffmpeg.get_ffmpeg_exe()
+	cmd = [ffmpeg_exe, "-i", str(path)]
+	proc = subprocess.run(cmd, stderr=subprocess.PIPE, text=True)
+	text = proc.stderr
+	
+	# Example parse: "Stream #0:0: Audio: mp3, 44100 Hz, stereo, ..."
+	m = re.search(r'Audio:.*?(\d+)\s*Hz.*?(mono|stereo)', text)
+	if not m:
+		raise RuntimeError("Could not parse audio info")
+	sr = int(m.group(1))
+	channels = 1 if m.group(2) == "mono" else 2
+	return sr, channels
 
-	.. code-block:: python
-		
-		import modusa as ms
-		audio_fp = ms.load(
-			"https://www.youtube.com/watch?v=lIpw9-Y_N0g",
-			sr = None, clip=(5, 10))
+def _load_audio_from_youtube(url: str):
+	"""
+	Download audio from a YouTube URL, convert it to WAV, and return the path.
+
+	Parameters
+	----------
+	url : str
+		YouTube video URL.
+
+	Returns
+	-------
+	Path
+		Path to the converted WAV file (you can delete it later).
+	"""
+	from modusa.tools.youtube_downloader import download
+	from modusa.tools.audio_converter import convert
+	import tempfile
+	
+	# Temporary directory to hold files (auto-created, not auto-deleted)
+	tmpdir = Path(tempfile.mkdtemp())
+	
+	# Download YouTube audio (e.g. .m4a or .webm)
+	audio_fp: Path = download(url=url, content_type="audio", output_dir=tmpdir)
+	
+	# Convert downloaded file to .wav
+	wav_audio_fp: Path = convert(inp_audio_fp=audio_fp, output_audio_fp=audio_fp.with_suffix(".wav"))
+	
+	# Return path to the WAV file
+	return wav_audio_fp
+
+#---------------------
+# Main Function
+#---------------------
+def load(path, sr=None, trim=None, ch=None):
+	"""
+	Lightweight audio loader using imageio-ffmpeg.
 
 	Parameters
 	----------
-	path: str
-		- Path to the audio
-		- Youtube URL
-	sr: int | None
+	path: PathLike/str/URL
+		- Path to the audio file / YouTube video
+	sr: int
 		- Sampling rate to load the audio in.
-	clip: number | tuple[number, number] | None
-		- Which segment of the audio you want.
-		- Eg., 10 => First 10 sec, (5, 10) => 5 to 10 second
-		- Default: None => Entire audio.
+		- Default: None => Use the original sampling rate
+	trim: tuple[number, number]
+		- (start, end) in seconds to trim the audio clip.
+		- Default: None => No trimming
+	ch: int
+		- 1 for mono and 2 for stereo
+		- Default: None => Use the original number of channels.
 
-	Return
-	------
+	Returns
+	-------
 	np.ndarray
-		- Audio signal.
-	int
-		- Sampling rate of the loaded audio signal.
-	title
-		- Title of the loaded audio.
-		- Filename without extension or YouTube title.
+		- Audio signal Float32 waveform in [-1, 1].
+	int:
+		Sampling rate.
+	str:
+		File name stem.
 	"""
+	path = Path(path)
+	ffmpeg_exe = ffmpeg.get_ffmpeg_exe()
 	
-	# Check if the path is YouTube
-	if ".youtube." in str(path):
-		# Download the audio in temp directory using tempfile module
-		with tempfile.TemporaryDirectory() as tmpdir:
-			# Download
-			audio_fp: Path = download(url=path, content_type="audio", output_dir=Path(tmpdir))
-			
-			# Convert the audio to ".wav" form for loading
-			wav_audio_fp: Path = convert(inp_audio_fp=audio_fp, output_audio_fp=audio_fp.with_suffix(".wav"))
-			
-			# Load the audio in memory
-			audio_data, audio_sr = sf.read(wav_audio_fp)
-			title = audio_fp.stem
-			
-			# Convert to mono if it's multi-channel
-			if audio_data.ndim > 1:
-				audio_data = audio_data.mean(axis=1)
-				
-			# Resample if needed
-			if sr is not None:
-				if audio_sr != sr:
-					n_samples = int(len(audio_data) * sr / audio_sr)
-					audio_data = resample(audio_data, n_samples)
-					audio_sr = sr
-		
-	else:
-		# Check if the file exists
-		fp = Path(path)
+	yt = False # Is the path a youtube URL
+	
+	if ".youtube" in str(path):
+		yt = True
+		try:
+			path: Path = _load_audio_from_youtube(url=str(path))
+		except Exception as e:
+			raise ConnectionRefusedError("unable to download from YouTube")
+	
+	# Find the real sample rate from the file
+	if sr is None:
+		sr, _ = _get_audio_info_ffmpeg(path)
+		if not (sr > 100 and sr < 80000):
+			raise Exception(f"Error reading the metadata for original sampling rate {sr}, please set `sr` explicitly")
+	
+	# Find the real number of channels from the file
+	if ch is None:
+		_, ch = _get_audio_info_ffmpeg(path)
 		
-		if not fp.exists():
-			raise FileNotFoundError(f"{path} does not exist.")
+		if ch not in [1, 2]:
+			raise Exception(f"Error reading the metadata for number of channels {ch}, please set `ch` explicitly")
 		
-		# Load the audio in memory
-		audio_data, audio_sr = sf.read(fp)
-		title = fp.stem
+	cmd = [ffmpeg_exe]
+	
+	# Optional trimming
+	if trim is not None:
+		start, end = trim
+		duration = end - start
+		cmd += ["-ss", str(start), "-t", str(duration)]
 		
-		# Convert to mono if it's multi-channel
-		if audio_data.ndim > 1:
-			audio_data = audio_data.mean(axis=1)
-			
-		# Resample if needed
-		if sr is not None:
-			if audio_sr != sr:
-				n_samples = int(len(audio_data) * sr / audio_sr)
-				audio_data = resample(audio_data, n_samples)
-				audio_sr = sr
-				
-	# Clip the audio signal as per needed
-	if clip is not None:
-		# Map clip input to the right format
-		if isinstance(clip, int or float):
-			clip = (0, clip)
-		elif isinstance(clip, tuple) and len(clip) > 1:
-			clip = (clip[0], clip[1])
-		else:
-			raise ValueError(f"Invalid clip type or length: {type(clip)}, len={len(clip)}")
+	cmd += ["-i", str(path), "-f", "s16le", "-acodec", "pcm_s16le"]
+	cmd += ["-ar", str(sr)]
+	cmd += ["-ac", str(ch)]
 		
-		audio_data = audio_data[int(clip[0]*audio_sr):int(clip[1]*audio_sr)]
+	cmd += ["-"]
+	
+	proc = subprocess.Popen(cmd, stdout=subprocess.PIPE, stderr=subprocess.DEVNULL)
+	raw = proc.stdout.read()
+	proc.wait()
 	
-	return audio_data, audio_sr, title
+	audio = np.frombuffer(raw, np.int16).astype(np.float32) / 32768.0
+	
+	# Stereo reshaping if forced
+	if ch == 2:
+		audio = audio.reshape(-1, 2).T
+		
+	# Delete the file if downloaded from youtube
+	if yt:
+		path.unlink(missing_ok=True)
+		path.parent.rmdir()
+		
+	return audio, sr, path.stem

modusa/tools/audio_player.py

@@ -5,17 +5,27 @@ from pathlib import Path
 import numpy as np
 import base64
 
-def play(
-	y: np.ndarray,
-	sr: float,
-	clip: tuple[float, float] | None = None,
-	label: str | None = None,
-) -> None:
+def play(y, sr: float, clip=None, label=None):
 	"""
-	Audio player with optional clip selection, transcription-style label,
-	and an embedded bottom-right logo (../images/icon.png).
+	Audio player with optional clip selection, transcription-style label.
+
+	Parameters
+	----------
+	y: ndarray
+		- Audio signal.
+	sr: float
+		- Sampling rate.
+	clip: tuple[float, float] | None
+		- The portion from the audio signal to be played.
+	label: str | None
+		- Could be transcription/labels attached to the audio.
+	
+	Returns
+	-------
+	None
 	"""
-	start_time, end_time = 0.0, len(y) / sr
+	start_time = 0.0
+	end_time = len(y) / sr if y.ndim < 2 else y[0].size / sr
 	
 	# Optional clip selection
 	if clip is not None:
@@ -84,7 +94,7 @@ def play(
 		box-shadow:0 1px 3px rgba(0,0,0,0.05);
 	">
 		{label_html}
-		<div style="margin-top:10px;">
+		<div style="margin-top:10px; margin-bottom:10px">
 			{audio_html}
 		</div>
 		{logo_html}

modusa/tools/plotter.py

@@ -209,6 +209,10 @@ class Fig:
 		
 		axs[-1, 0].tick_params(bottom=True, labelbottom=True)
 		
+		# Add the figure title on top-left (if any)
+		if self._fig_num is not None:
+			fig.suptitle(f'fig - {self._fig_num}', fontsize=12, fontweight='bold', x=0.01, ha='left', va='top', y=0.98)
+			
 		# xlim should be applied on reference subplot, rest all subplots will automatically adjust
 		if xlim is not None:
 			axs[0, 0].set_xlim(xlim)

{modusa-0.4.20.dist-info → modusa-0.4.23.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: modusa
-Version: 0.4.20
+Version: 0.4.23
 Summary: A modular signal analysis python library.
 Author-Email: Ankit Anand <ankit0.anand0@gmail.com>
 License: MIT
@@ -9,10 +9,8 @@ Requires-Dist: numpy>=2.2.6
 Requires-Dist: matplotlib>=3.10.3
 Requires-Dist: yt-dlp==2025.9.23
 Requires-Dist: IPython>=9.5.0
-Requires-Dist: sounddevice>=0.5.2
 Requires-Dist: ipywidgets>=8.1.7
-Requires-Dist: soundfile>=0.13.1
-Requires-Dist: scipy>=1.16.2
+Requires-Dist: imageio-ffmpeg>=0.6.0
 Description-Content-Type: text/markdown
 
 # modusa

{modusa-0.4.20.dist-info → modusa-0.4.23.dist-info}/RECORD

@@ -1,8 +1,9 @@
-modusa-0.4.20.dist-info/METADATA,sha256=pJ932mbsVsJ8XTbIHPX3h1RZa5k-GUBk0i5zgsqGMKY,1467
-modusa-0.4.20.dist-info/WHEEL,sha256=9P2ygRxDrTJz3gsagc0Z96ukrxjr-LFBGOgv3AuKlCA,90
-modusa-0.4.20.dist-info/entry_points.txt,sha256=fmKpleVXj6CdaBVL14WoEy6xx7JQCs85jvzwTi3lePM,73
-modusa-0.4.20.dist-info/licenses/LICENSE.md,sha256=JTaXAjx5awk76VArKCx5dUW8vmLEWsL_ZlR7-umaHbA,1078
-modusa/__init__.py,sha256=x95HzUy8e7c6XeEIZsuy6QmE82fcfJsnWvTUm8u6WyE,324
+modusa-0.4.23.dist-info/METADATA,sha256=VcG0sWHvNgl2n0jMCiKiWM2BAeGHUJorSp0pFogxx90,1408
+modusa-0.4.23.dist-info/WHEEL,sha256=9P2ygRxDrTJz3gsagc0Z96ukrxjr-LFBGOgv3AuKlCA,90
+modusa-0.4.23.dist-info/entry_points.txt,sha256=fmKpleVXj6CdaBVL14WoEy6xx7JQCs85jvzwTi3lePM,73
+modusa-0.4.23.dist-info/licenses/LICENSE.md,sha256=JTaXAjx5awk76VArKCx5dUW8vmLEWsL_ZlR7-umaHbA,1078
+modusa/.DS_Store,sha256=hHecruWmea5u7ITMSdq62sVVdvT2I5lTN20gTfD3Btk,8196
+modusa/__init__.py,sha256=90isCiArafICPkaTgChbljh2F5o_iDKhqMz1gArlDOE,324
 modusa/config.py,sha256=bTqK4t00FZqERVITrxW_q284aDDJAa9aMSfFknfR-oU,280
 modusa/decorators.py,sha256=8zeNX_wE37O6Vp0ysR4-WCZaEL8mq8dyCF_I5DHOzks,5905
 modusa/devtools/generate_docs_source.py,sha256=UDflHsk-Yh9-3YJTVBzKL32y8hcxiRgAlFEBTMiDqwM,3301
@@ -42,15 +43,15 @@ modusa/plugins/__init__.py,sha256=r1Bf5mnrVKRIwxboutY1iGzDy4EPQhqpk1kSW7iJj_Q,54
 modusa/plugins/base.py,sha256=Bh_1Bja7fOymFsCgwhXDbV6ys3D8muNrPwrfDrG_G_A,2382
 modusa/tools/__init__.py,sha256=S7P1uYckyUdkha2UX9uj4P7mqpF6cc5SHqiW6NEupgs,342
 modusa/tools/_plotter_old.py,sha256=KGow7mihA2H1WNq7s5bpivhCgGo2qVIeDaO6iabpsrg,19294
-modusa/tools/ann_loader.py,sha256=BEdwAh_lccx_SnAO3bNMeY3O5zGiJlH2o4snWmXj8eQ,3034
+modusa/tools/ann_loader.py,sha256=m6Qu6jXnQ8LfUhKItoHSaHlGxUyzUJlGEyu4_50qJ8w,3099
 modusa/tools/audio_converter.py,sha256=415qBoPm2sBIuBSI7m1XBKm0AbmVmPydIPPr-uO8D3c,1778
-modusa/tools/audio_loader.py,sha256=n9Q9t_GmlE8AtioVwRcXX3rnd6PkbGTx-hAoNgUnNOQ,2780
-modusa/tools/audio_player.py,sha256=BI1ZmETxnmJACDHZMsbplzspwAx-_oKe3SpaHz-MFds,2295
+modusa/tools/audio_loader.py,sha256=nQl-E8xM1wdoYzZ3yGw25FJY8EogbqIZzPSMFt4Fv1E,3899
+modusa/tools/audio_player.py,sha256=kyBUnodkOE9Ox-hKHkfPeGAQ1RPTddbZYXO1ezz6-9w,2494
 modusa/tools/audio_recorder.py,sha256=K_LGqsPdjTdf3figEZTSQLmgMzYWgz18HTO8C1j5fE4,2788
 modusa/tools/audio_saver.py,sha256=ldzfr_AydsHTnKbxmBLJblN-hLzTmOlppOm306xI4Ug,510
 modusa/tools/base.py,sha256=C0ESJ0mIfjjRlAkRbSetNtMoOfS6IrHBjexRp3l_Mh4,1293
 modusa/tools/math_ops.py,sha256=ZZ7U4DgqT7cOeE7_Lzi_Qq-48WYfwR9_osbZwTmE9eg,8690
-modusa/tools/plotter.py,sha256=V0CcRnXkE4tkTZQe6v0LI0ucEagqeudOZ-xoQw2clTc,30667
+modusa/tools/plotter.py,sha256=RrELZYgnyUVhGzcNfpTvyZvVTmZ6dvQmN9YSQoUjmaM,30860
 modusa/tools/youtube_downloader.py,sha256=hB_X8-7nOHXOlxg6vv3wyhBLoAsWyomrULP6_uCQL7s,1698
 modusa/utils/__init__.py,sha256=1oLL20yLB1GL9IbFiZD8OReDqiCpFr-yetIR6x1cNkI,23
 modusa/utils/config.py,sha256=cuGbqbovx5WDQq5rw3hIKcv3CnE5NttjacSOWnP1yu4,576
@@ -58,4 +59,4 @@ modusa/utils/excp.py,sha256=L9vhaGjKpv9viJYdmC9n5ndmk2GVbUBuFyZyhAQZmWY,906
 modusa/utils/logger.py,sha256=K0rsnObeNKCxlNeSnVnJeRhgfmob6riB2uyU7h3dDmA,571
 modusa/utils/np_func_cat.py,sha256=TyIFgRc6bARRMDnZxlVURO5Z0I-GWhxRONYyIv-Vwxs,1007
 modusa/utils/plot.py,sha256=s_vNdxvKfwxEngvJPgrF1PcmxZNnNaaXPViHWjyjJ-c,5335
-modusa-0.4.20.dist-info/RECORD,,
+modusa-0.4.23.dist-info/RECORD,,