auto-editor 24.29.1__py3-none-any.whl → 24.31.1__py3-none-any.whl

auto_editor/__init__.py

@@ -1,2 +1,2 @@
-__version__ = "24.29.1"
-version = "24w29a"
+__version__ = "24.31.1"
+version = "24w31a"

auto_editor/analyze.py

@@ -4,23 +4,25 @@ import os
 import re
 from dataclasses import dataclass
 from fractions import Fraction
+from math import ceil
 from typing import TYPE_CHECKING
 
+import av
 import numpy as np
+from av.audio.fifo import AudioFifo
+from av.subtitles.subtitle import AssSubtitle
 
 from auto_editor import version
 from auto_editor.utils.subtitle_tools import convert_ass_to_text
-from auto_editor.wavfile import read
 
 if TYPE_CHECKING:
+    from collections.abc import Iterator
     from fractions import Fraction
     from typing import Any
 
-    from av.filter import FilterContext
     from numpy.typing import NDArray
 
     from auto_editor.ffwrapper import FileInfo
-    from auto_editor.output import Ensure
     from auto_editor.utils.bar import Bar
     from auto_editor.utils.log import Log
 
@@ -28,7 +30,6 @@ if TYPE_CHECKING:
 @dataclass(slots=True)
 class FileSetup:
     src: FileInfo
-    ensure: Ensure
     strict: bool
     tb: Fraction
     bar: Bar
@@ -40,11 +41,6 @@ class LevelError(Exception):
     pass
 
 
-def link_nodes(*nodes: FilterContext) -> None:
-    for c, n in zip(nodes, nodes[1:]):
-        c.link_to(n)
-
-
 def mut_remove_small(
     arr: NDArray[np.bool_], lim: int, replace: int, with_: int
 ) -> None:
@@ -92,9 +88,90 @@ def obj_tag(tag: str, tb: Fraction, obj: dict[str, Any]) -> str:
     return key
 
 
+def iter_audio(src, tb: Fraction, stream: int = 0) -> Iterator[float]:
+    fifo = AudioFifo()
+    try:
+        container = av.open(src.path, "r")
+        audio_stream = container.streams.audio[stream]
+        sample_rate = audio_stream.rate
+
+        exact_size = (1 / tb) * sample_rate
+        accumulated_error = 0
+
+        # Resample so that audio data is between [-1, 1]
+        resampler = av.AudioResampler(
+            av.AudioFormat("flt"), audio_stream.layout, sample_rate
+        )
+
+        for frame in container.decode(audio=stream):
+            frame.pts = None  # Skip time checks
+
+            for reframe in resampler.resample(frame):
+                fifo.write(reframe)
+
+            while fifo.samples >= ceil(exact_size):
+                size_with_error = exact_size + accumulated_error
+                current_size = round(size_with_error)
+                accumulated_error = size_with_error - current_size
+
+                audio_chunk = fifo.read(current_size)
+                assert audio_chunk is not None
+                arr = audio_chunk.to_ndarray().flatten()
+                yield float(np.max(np.abs(arr)))
+
+    finally:
+        container.close()
+
+
+def iter_motion(src, tb, stream: int, blur: int, width: int) -> Iterator[float]:
+    container = av.open(src.path, "r")
+
+    video = container.streams.video[stream]
+    video.thread_type = "AUTO"
+
+    prev_frame = None
+    current_frame = None
+    total_pixels = src.videos[0].width * src.videos[0].height
+    index = 0
+    prev_index = -1
+
+    graph = av.filter.Graph()
+    graph.link_nodes(
+        graph.add_buffer(template=video),
+        graph.add("scale", f"{width}:-1"),
+        graph.add("format", "gray"),
+        graph.add("gblur", f"sigma={blur}"),
+        graph.add("buffersink"),
+    ).configure()
+
+    for unframe in container.decode(video):
+        if unframe.pts is None:
+            continue
+
+        graph.push(unframe)
+        frame = graph.pull()
+        assert frame.time is not None
+        index = round(frame.time * tb)
+
+        current_frame = frame.to_ndarray()
+        if prev_frame is None:
+            value = 0.0
+        else:
+            # Use `int16` to avoid underflow with `uint8` datatype
+            diff = np.abs(prev_frame.astype(np.int16) - current_frame.astype(np.int16))
+            value = np.count_nonzero(diff) / total_pixels
+
+        for _ in range(index - prev_index):
+            yield value
+
+        prev_frame = current_frame
+        prev_index = index
+
+    container.close()
+
+
 @dataclass(slots=True)
 class Levels:
-    ensure: Ensure
     src: FileInfo
     tb: Fraction
     bar: Bar
@@ -107,26 +184,16 @@ class Levels:
             if (arr := self.read_cache("audio", {"stream": 0})) is not None:
                 return len(arr)
 
-            sr, samples = read(self.ensure.audio(self.src, 0))
-            samp_count = len(samples)
-            del samples
-
-            samp_per_ticks = sr / self.tb
-            ticks = int(samp_count / samp_per_ticks)
-            self.log.debug(f"Audio Length: {ticks}")
-            self.log.debug(
-                f"... without rounding: {float(samp_count / samp_per_ticks)}"
-            )
-            return ticks
+            result = sum(1 for _ in iter_audio(self.src, self.tb, 0))
+            self.log.debug(f"Audio Length: {result}")
+            return result
 
         # If there's no audio, get length in video metadata.
-        import av
-
-        with av.open(f"{self.src.path}") as cn:
-            if len(cn.streams.video) < 1:
+        with av.open(self.src.path) as container:
+            if len(container.streams.video) == 0:
                 self.log.error("Could not get media duration")
 
-            video = cn.streams.video[0]
+            video = container.streams.video[0]
 
             if video.duration is None or video.time_base is None:
                 dur = 0
@@ -171,56 +238,70 @@ class Levels:
         return arr
 
     def audio(self, stream: int) -> NDArray[np.float64]:
-        if stream > len(self.src.audios) - 1:
+        if stream >= len(self.src.audios):
             raise LevelError(f"audio: audio stream '{stream}' does not exist.")
 
         if (arr := self.read_cache("audio", {"stream": stream})) is not None:
             return arr
 
-        sr, samples = read(self.ensure.audio(self.src, stream))
-
-        if len(samples) == 0:
-            raise LevelError(f"audio: stream '{stream}' has no samples.")
-
-        def get_max_volume(s: np.ndarray) -> float:
-            return max(float(np.max(s)), -float(np.min(s)))
-
-        max_volume = get_max_volume(samples)
-        self.log.debug(f"Max volume: {max_volume}")
+        with av.open(self.src.path, "r") as container:
+            audio = container.streams.audio[stream]
+            if audio.duration is not None and audio.time_base is not None:
+                inaccurate_dur = int(audio.duration * audio.time_base * self.tb)
+            elif container.duration is not None:
+                inaccurate_dur = int(container.duration / av.time_base * self.tb)
+            else:
+                inaccurate_dur = 1024
 
-        samp_count = samples.shape[0]
-        samp_per_ticks = sr / self.tb
+        bar = self.bar
+        bar.start(inaccurate_dur, "Analyzing audio volume")
 
-        if samp_per_ticks < 1:
-            self.log.error(
-                f"audio: stream '{stream}'\n  Samplerate ({sr}) must be greater than "
-                f"or equal to timebase ({self.tb})\n"
-                "  Try `-fps 30` and/or `--sample-rate 48000`"
-            )
+        result = np.zeros((inaccurate_dur), dtype=np.float64)
+        index = 0
+        for value in iter_audio(self.src, self.tb, stream):
+            if index > len(result) - 1:
+                result = np.concatenate(
+                    (result, np.zeros((len(result)), dtype=np.float64))
+                )
+            result[index] = value
+            bar.tick(index)
+            index += 1
 
-        audio_ticks = int(samp_count / samp_per_ticks)
-        self.log.debug(
-            f"analyze: audio length: {audio_ticks} ({float(samp_count / samp_per_ticks)})"
-        )
-        self.bar.start(audio_ticks, "Analyzing audio volume")
+        bar.end()
+        return self.cache("audio", {"stream": stream}, result[:index])
 
-        threshold_list = np.zeros((audio_ticks), dtype=np.float64)
+    def motion(self, stream: int, blur: int, width: int) -> NDArray[np.float64]:
+        if stream >= len(self.src.videos):
+            raise LevelError(f"motion: video stream '{stream}' does not exist.")
 
-        if max_volume == 0:  # Prevent dividing by zero
-            return threshold_list
+        mobj = {"stream": stream, "width": width, "blur": blur}
+        if (arr := self.read_cache("motion", mobj)) is not None:
+            return arr
 
-        # Determine when audio is silent or loud.
-        for i in range(audio_ticks):
-            if i % 500 == 0:
-                self.bar.tick(i)
+        with av.open(self.src.path, "r") as container:
+            video = container.streams.video[stream]
+            inaccurate_dur = (
+                1024
+                if video.duration is None or video.time_base is None
+                else int(video.duration * video.time_base * self.tb)
+            )
 
-            start = int(i * samp_per_ticks)
-            end = min(int((i + 1) * samp_per_ticks), samp_count)
+        bar = self.bar
+        bar.start(inaccurate_dur, "Analyzing motion")
 
-            threshold_list[i] = get_max_volume(samples[start:end]) / max_volume
+        result = np.zeros((inaccurate_dur), dtype=np.float64)
+        index = 0
+        for value in iter_motion(self.src, self.tb, stream, blur, width):
+            if index > len(result) - 1:
+                result = np.concatenate(
+                    (result, np.zeros((len(result)), dtype=np.float64))
+                )
+            result[index] = value
+            bar.tick(index)
+            index += 1
 
-        self.bar.end()
-        return self.cache("audio", {"stream": stream}, threshold_list)
+        bar.end()
+        return self.cache("motion", mobj, result[:index])
 
     def subtitle(
         self,
@@ -238,9 +319,6 @@ class Levels:
         except re.error as e:
             self.log.error(e)
 
-        import av
-        from av.subtitles.subtitle import AssSubtitle, TextSubtitle
-
         try:
             container = av.open(self.src.path, "r")
             subtitle_stream = container.streams.subtitles[stream]
@@ -287,8 +365,6 @@ class Levels:
                 for sub in subset:
                     if isinstance(sub, AssSubtitle):
                         line = convert_ass_to_text(sub.ass.decode(errors="ignore"))
-                    elif isinstance(sub, TextSubtitle):
-                        line = sub.text.decode(errors="ignore")
                     else:
                         continue
 
@@ -299,68 +375,3 @@ class Levels:
         container.close()
 
         return result
-
-    def motion(self, stream: int, blur: int, width: int) -> NDArray[np.float64]:
-        import av
-
-        if stream >= len(self.src.videos):
-            raise LevelError(f"motion: video stream '{stream}' does not exist.")
-
-        mobj = {"stream": stream, "width": width, "blur": blur}
-        if (arr := self.read_cache("motion", mobj)) is not None:
-            return arr
-
-        container = av.open(f"{self.src.path}", "r")
-
-        video = container.streams.video[stream]
-        video.thread_type = "AUTO"
-
-        inaccurate_dur = 1 if video.duration is None else video.duration
-        self.bar.start(inaccurate_dur, "Analyzing motion")
-
-        prev_frame = None
-        current_frame = None
-        total_pixels = self.src.videos[0].width * self.src.videos[0].height
-        index = 0
-
-        graph = av.filter.Graph()
-        link_nodes(
-            graph.add_buffer(template=video),
-            graph.add("scale", f"{width}:-1"),
-            graph.add("format", "gray"),
-            graph.add("gblur", f"sigma={blur}"),
-            graph.add("buffersink"),
-        )
-        graph.configure()
-
-        threshold_list = np.zeros((1024), dtype=np.float64)
-
-        for unframe in container.decode(video):
-            graph.push(unframe)
-            frame = graph.pull()
-
-            # Showing progress ...
-            assert frame.time is not None
-            index = int(frame.time * self.tb)
-            if frame.pts is not None:
-                self.bar.tick(frame.pts)
-
-            current_frame = frame.to_ndarray()
-
-            if index > len(threshold_list) - 1:
-                threshold_list = np.concatenate(
-                    (threshold_list, np.zeros((len(threshold_list)), dtype=np.float64)),
-                    axis=0,
-                )
-
-            if prev_frame is not None:
-                # Use `int16` to avoid underflow with `uint8` datatype
-                diff = np.abs(
-                    prev_frame.astype(np.int16) - current_frame.astype(np.int16)
-                )
-                threshold_list[index] = np.count_nonzero(diff) / total_pixels
-
-            prev_frame = current_frame
-
-        self.bar.end()
-        return self.cache("motion", mobj, threshold_list[:index])

auto_editor/edit.py

@@ -68,12 +68,8 @@ def set_video_codec(
 ) -> str:
     if codec == "auto":
         codec = "h264" if (src is None or not src.videos) else src.videos[0].codec
-        if ctr.vcodecs is not None:
-            if ctr.vstrict and codec not in ctr.vcodecs:
-                return ctr.vcodecs[0]
-
-            if codec in ctr.disallow_v:
-                return ctr.vcodecs[0]
+        if codec not in ctr.vcodecs and ctr.default_vid != "none":
+            return ctr.default_vid
         return codec
 
     if codec == "copy":
@@ -83,12 +79,7 @@ def set_video_codec(
             log.error("Input file does not have a video stream to copy codec from.")
         codec = src.videos[0].codec
 
-    if ctr.vstrict:
-        assert ctr.vcodecs is not None
-        if codec not in ctr.vcodecs:
-            log.error(codec_error.format(codec, out_ext))
-
-    if codec in ctr.disallow_v:
+    if ctr.vcodecs is not None and codec not in ctr.vcodecs:
         log.error(codec_error.format(codec, out_ext))
 
     return codec
@@ -99,8 +90,10 @@ def set_audio_codec(
 ) -> str:
     if codec == "auto":
         codec = "aac" if (src is None or not src.audios) else src.audios[0].codec
-        if ctr.acodecs is not None and codec not in ctr.acodecs:
-            return ctr.acodecs[0]
+        if codec not in ctr.acodecs and ctr.default_aud != "none":
+            return ctr.default_aud
+        if codec == "mp3float":
+            return "mp3"
         return codec
 
     if codec == "copy":
@@ -209,10 +202,8 @@ def edit_media(
     else:
         samplerate = args.sample_rate
 
-    ensure = Ensure(ffmpeg, bar, samplerate, temp, log)
-
     if tl is None:
-        tl = make_timeline(sources, ensure, args, samplerate, bar, temp, log)
+        tl = make_timeline(sources, args, samplerate, bar, temp, log)
 
     if export["export"] == "timeline":
         from auto_editor.formats.json import make_json_timeline
@@ -223,7 +214,7 @@ def edit_media(
     if args.preview:
         from auto_editor.preview import preview
 
-        preview(ensure, tl, temp, log)
+        preview(tl, temp, log)
         return
 
     if export["export"] == "json":
@@ -272,13 +263,15 @@ def edit_media(
         sub_output = []
         apply_later = False
 
-        if ctr.allow_subtitle and not args.sn:
+        ensure = Ensure(ffmpeg, bar, samplerate, temp, log)
+
+        if ctr.default_sub != "none" and not args.sn:
             sub_output = make_new_subtitles(tl, ensure, temp)
 
-        if ctr.allow_audio:
+        if ctr.default_aud != "none":
             audio_output = make_new_audio(tl, ensure, args, ffmpeg, bar, temp, log)
 
-        if ctr.allow_video:
+        if ctr.default_vid != "none":
             if tl.v:
                 out_path, apply_later = render_av(ffmpeg, tl, args, bar, ctr, temp, log)
                 visual_output.append((True, out_path))

auto_editor/lang/palet.py

@@ -620,42 +620,36 @@ def make_array(dtype: Sym, size: int, v: int = 0) -> np.ndarray:
         raise MyError(f"number too large to be converted to {dtype}")
 
 
-def minclip(oarr: BoolList, _min: int) -> BoolList:
+def minclip(oarr: BoolList, _min: int, /) -> BoolList:
     arr = np.copy(oarr)
     mut_remove_small(arr, _min, replace=1, with_=0)
     return arr
 
 
-def mincut(oarr: BoolList, _min: int) -> BoolList:
+def mincut(oarr: BoolList, _min: int, /) -> BoolList:
     arr = np.copy(oarr)
     mut_remove_small(arr, _min, replace=0, with_=1)
     return arr
 
 
-def maxclip(oarr: BoolList, _min: int) -> BoolList:
+def maxclip(oarr: BoolList, _min: int, /) -> BoolList:
     arr = np.copy(oarr)
     mut_remove_large(arr, _min, replace=1, with_=0)
     return arr
 
 
-def maxcut(oarr: BoolList, _min: int) -> BoolList:
+def maxcut(oarr: BoolList, _min: int, /) -> BoolList:
     arr = np.copy(oarr)
     mut_remove_large(arr, _min, replace=0, with_=1)
     return arr
 
 
-def margin(a: int, b: Any, c: Any = None) -> BoolList:
-    if c is None:
-        check_args("margin", [a, b], (2, 2), (is_int, is_boolarr))
-        oarr = b
-        start, end = a, a
-    else:
-        check_args("margin", [a, b, c], (3, 3), (is_int, is_int, is_boolarr))
-        oarr = c
-        start, end = a, b
-
+def margin(oarr: BoolList, start: int, end: int | None = None, /) -> BoolList:
     arr = np.copy(oarr)
-    mut_margin(arr, start, end)
+    if end is None:
+        mut_margin(arr, start, start)
+    else:
+        mut_margin(arr, start, end)
     return arr
 
 
@@ -1741,6 +1735,8 @@ env.update({
     "round": Proc("round", round, (1, 1), is_real),
     "max": Proc("max", lambda *v: max(v), (1, None), is_real),
     "min": Proc("min", lambda *v: min(v), (1, None), is_real),
+    "max-seq": Proc("max-seq", max, (1, 1), is_sequence),
+    "min-seq": Proc("min-seq", min, (1, 1), is_sequence),
     "mod": Proc("mod", mod, (2, 2), is_int),
     "modulo": Proc("modulo", mod, (2, 2), is_int),
     # symbols
@@ -1796,7 +1792,7 @@ env.update({
     "bool-array": Proc(
         "bool-array", lambda *a: np.array(a, dtype=np.bool_), (1, None), is_nat
     ),
-    "margin": Proc("margin", margin, (2, 3)),
+    "margin": Proc("margin", margin, (2, 3), is_boolarr, is_int),
     "mincut": Proc("mincut", mincut, (2, 2), is_boolarr, is_nat),
     "minclip": Proc("minclip", minclip, (2, 2), is_boolarr, is_nat),
     "maxcut": Proc("maxcut", maxcut, (2, 2), is_boolarr, is_nat),

auto_editor/lib/data_structs.py

@@ -185,6 +185,8 @@ def display_str(val: object) -> str:
         return f"{val.real}{join}{val.imag}i"
     if type(val) is np.bool_:
         return "1" if val else "0"
+    if type(val) is np.float64 or type(val) is np.float32:
+        return f"{float(val)}"
     if type(val) is Fraction:
         return f"{val.numerator}/{val.denominator}"
 

auto_editor/make_layers.py

@@ -18,7 +18,6 @@ from auto_editor.utils.types import Args, CoerceError, time
 if TYPE_CHECKING:
     from numpy.typing import NDArray
 
-    from auto_editor.output import Ensure
     from auto_editor.utils.bar import Bar
     from auto_editor.utils.chunks import Chunks
     from auto_editor.utils.log import Log
@@ -75,7 +74,6 @@ def make_av(src: FileInfo, all_clips: list[list[Clip]]) -> tuple[VSpace, ASpace]
 def run_interpreter_for_edit_option(
     text: str, filesetup: FileSetup
 ) -> NDArray[np.bool_]:
-    ensure = filesetup.ensure
     src = filesetup.src
     tb = filesetup.tb
     bar = filesetup.bar
@@ -87,8 +85,8 @@ def run_interpreter_for_edit_option(
         if log.is_debug:
             log.debug(f"edit: {parser}")
 
-        env["timebase"] = filesetup.tb
-        env["@levels"] = Levels(ensure, src, tb, bar, temp, log)
+        env["timebase"] = tb
+        env["@levels"] = Levels(src, tb, bar, temp, log)
         env["@filesetup"] = filesetup
 
         results = interpret(env, parser)
@@ -139,7 +137,6 @@ def parse_time(val: str, arr: NDArray, tb: Fraction) -> int:  # raises: `CoerceE
 
 def make_timeline(
     sources: list[FileInfo],
-    ensure: Ensure,
     args: Args,
     sr: int,
     bar: Bar,
@@ -169,7 +166,7 @@ def make_timeline(
     concat = np.concatenate
 
     for i, src in enumerate(sources):
-        filesetup = FileSetup(src, ensure, len(sources) < 2, tb, bar, temp, log)
+        filesetup = FileSetup(src, len(sources) < 2, tb, bar, temp, log)
 
         edit_result = run_interpreter_for_edit_option(method, filesetup)
         mut_margin(edit_result, start_margin, end_margin)

auto_editor/output.py

@@ -57,7 +57,7 @@ class Ensure:
             output_astream = out_container.add_stream("pcm_s16le", rate=sample_rate)
             assert isinstance(output_astream, av.audio.stream.AudioStream)
 
-            resampler = AudioResampler(format="s16", layout="stereo", rate=sample_rate)  # type: ignore
+            resampler = AudioResampler(format="s16", layout="stereo", rate=sample_rate)
             for i, frame in enumerate(in_container.decode(astream)):
                 if i % 1500 == 0:
                     bar.tick(0 if frame.time is None else frame.time)
@@ -99,7 +99,7 @@ def _ffset(option: str, value: str | None) -> list[str]:
     return [option] + [value]
 
 
-def video_quality(args: Args, ctr: Container) -> list[str]:
+def video_quality(args: Args) -> list[str]:
     return (
         _ffset("-b:v", args.video_bitrate)
         + ["-c:v", args.video_codec]
@@ -174,7 +174,7 @@ def mux_quality_media(
     for is_video, path in visual_output:
         if is_video:
             if apply_v:
-                cmd += video_quality(args, ctr)
+                cmd += video_quality(args)
             else:
                 # Real video is only allowed on track 0
                 cmd += ["-c:v:0", "copy"]
@@ -211,7 +211,7 @@ def mux_quality_media(
             cmd.extend(["-c:s", scodec])
         elif ctr.scodecs is not None:
             if scodec not in ctr.scodecs:
-                scodec = ctr.scodecs[0]
+                scodec = ctr.default_sub
             cmd.extend(["-c:s", scodec])
 
     if a_tracks > 0:
@@ -227,7 +227,7 @@ def mux_quality_media(
             cmd.extend(["-color_range", f"{color_range}"])
         if colorspace in (0, 1) or (colorspace >= 3 and colorspace < 16):
             cmd.extend(["-colorspace", f"{colorspace}"])
-        if color_prim in (0, 1) or (color_prim >= 4 and color_prim < 17):
+        if color_prim == 1 or (color_prim >= 4 and color_prim < 17):
             cmd.extend(["-color_primaries", f"{color_prim}"])
         if color_trc == 1 or (color_trc >= 4 and color_trc < 22):
             cmd.extend(["-color_trc", f"{color_trc}"])

auto_editor/preview.py

@@ -6,7 +6,6 @@ from statistics import fmean, median
 from typing import TextIO
 
 from auto_editor.analyze import Levels
-from auto_editor.output import Ensure
 from auto_editor.timeline import v3
 from auto_editor.utils.bar import Bar
 from auto_editor.utils.func import to_timecode
@@ -49,7 +48,7 @@ def all_cuts(tl: v3, in_len: int) -> list[int]:
     return cut_lens
 
 
-def preview(ensure: Ensure, tl: v3, temp: str, log: Log) -> None:
+def preview(tl: v3, temp: str, log: Log) -> None:
     log.conwrite("")
     tb = tl.tb
 
@@ -66,7 +65,7 @@ def preview(ensure: Ensure, tl: v3, temp: str, log: Log) -> None:
 
     in_len = 0
     for src in all_sources:
-        in_len += Levels(ensure, src, tb, Bar("none"), temp, log).media_length
+        in_len += Levels(src, tb, Bar("none"), temp, log).media_length
 
     out_len = tl.out_len()
 

auto_editor/render/subtitle.py

@@ -49,7 +49,7 @@ class SubtitleParser:
         self.codec = codec
         self.contents = []
 
-        if codec == "ass":
+        if codec == "ass" or codec == "ssa":
             time_code = re.compile(r"(.*)(\d+:\d+:[\d.]+)(.*)(\d+:\d+:[\d.]+)(.*)")
         elif codec == "webvtt":
             time_code = re.compile(r"()(\d+:[\d.]+)( --> )(\d+:[\d.]+)(\n.*)")