audio AND video silence now drops

src-lib/testdata/poc.sh

@@ -6,48 +6,97 @@ lines = [i.strip() for i in stdin.readlines()]
 
 def frame_duplicate_status(line):
    return "keep" if "keep" in line else "drop", "Parsed_mpdecimate_0" in line
-somelines = []
+def silence_status(line):
+   if not "[silencedetect @ " in line:
+      return None
+   return line.split()[3:5]
+def parse_duration(line):
+   if not line.startswith("Duration: "):
+      return None
+   ts = line.split()[1].strip(",")
+   hours, minutes, seconds = [float(i) for i in ts.split(":")]
+   return seconds + 60.0 * minutes + 60.0 * 60.0 * hours
+
+state_transition_lines_video = []
+state_transition_lines_audio = [0.0]
+state_total_duration = 0.0
 for i in range(len(lines)):
    status, relevant = frame_duplicate_status(lines[i])
-  if not relevant:
+   if relevant:
-    continue
+      if not state_transition_lines_video:
-  if not somelines:
+         state_transition_lines_video.append(lines[i])
-    somelines.append(lines[i])
       else:
-    prev_status, _ = frame_duplicate_status(somelines[-1])
+         prev_status, _ = frame_duplicate_status(state_transition_lines_video[-1])
          if status != prev_status:
-      somelines.append(lines[i-1])
+            state_transition_lines_video.append(lines[i-1])
-      somelines.append(lines[i])
+            state_transition_lines_video.append(lines[i])
-somelines = [(frame_duplicate_status(i)[0], i.split("pts_time:")[1].split()[0]) for i in somelines]
+   status = silence_status(lines[i])
+   if status:
+      state_transition_lines_audio.append(float(status[1]))
+   status = parse_duration(lines[i])
+   if not status is None:
+      duration = status
+state_transition_lines_audio.append(duration)
+state_transition_lines_video = [(frame_duplicate_status(i)[0], i.split("pts_time:")[1].split()[0]) for i in state_transition_lines_video]
 
 # just the keeps
-somelines = [i for i in somelines[2:] if i[0] == "keep"]
+state_transition_lines_video = [i for i in state_transition_lines_video[2:] if i[0] == "keep"]
 
-# translate to floats
+# translate to float spans
-drops = []
+spans_to_keep_video = []
-for i in range(0, len(somelines), 2):
+spans_to_keep_audio = []
-   start = float(somelines[i][1])
+for i in range(0, len(state_transition_lines_video), 2):
-   if len(somelines) > i+1:
+   start = float(state_transition_lines_video[i][1])
-      stop = float(somelines[i+1][1])
+   if len(state_transition_lines_video) > i+1:
+      stop = float(state_transition_lines_video[i+1][1])
       if stop - start > 1:
-         drops.append((start, stop))
+         spans_to_keep_video.append((start, stop))
-   #else:
+   else:
-   #   drops.append((start, "inf"))
+      spans_to_keep_video.append((start, duration))
+spans_to_keep_audio = [
+   (state_transition_lines_audio[i], state_transition_lines_audio[i+1])
+   for i in range(0, len(state_transition_lines_audio), 2)
+]
+
+# drop zero len spans
+spans_to_keep_video = [i for i in spans_to_keep_video if i[1]-i[0] > 0.1]
+spans_to_keep_audio = [i for i in spans_to_keep_audio if i[1]-i[0] > 0.1]
 
 # extend around
-for i in range(len(drops)):
+for arr in [spans_to_keep_video, spans_to_keep_audio]:
-   drops[i] = (drops[i][0] - 1.0, drops[i][1] + 1.0)
+   for i in range(len(arr)):
+      arr[i] = (
+         max(0.0, arr[i][0] - 1.0),
+         min(duration, arr[i][1] + 1.0),
+      )
 
 # combine near
-for i in range(len(drops)-1, 0, -1):
+for arr in [spans_to_keep_video, spans_to_keep_audio]:
-   first_start, first_stop = drops[i-1]
+   for i in range(len(arr)-1, 0, -1):
-   secnd_start, secnd_stop = drops[i]
+      first_start, first_stop = arr[i-1]
+      secnd_start, secnd_stop = arr[i]
       if secnd_start - first_stop < 5:
-      drops = drops[:i-1] + [(first_start, secnd_stop)] + drops[i+1:]
+         arr = arr[:i-1] + [(first_start, secnd_stop)] + arr[i+1:]
+
+# union audio, video spans
+spans_to_keep = spans_to_keep_video + spans_to_keep_audio
+spans_to_keep = sorted(spans_to_keep)
+for i in range(len(spans_to_keep)-1, 0, -1):
+   first_start, first_stop = spans_to_keep[i-1]
+   secnd_start, secnd_stop = spans_to_keep[i]
+   if secnd_start - first_stop < 5:
+      spans_to_keep = spans_to_keep[:i-1] + [(first_start, secnd_stop)] + spans_to_keep[i+1:]
+
+print("\n".join([
+   f'AUDIO: {i[0]:.2f},{i[1]-i[0]:.2f}' for i in spans_to_keep_audio
+]), file=stderr)
+print("\n".join([
+   f'VIDEO: {i[0]:.2f},{i[1]-i[0]:.2f}' for i in spans_to_keep_video
+]), file=stderr)
 
 # output
 print("\n".join([
-   f'{i[0]:.2f},{i[1]-i[0]:.2f}' for i in drops
+   f'{i[0]:.2f},{i[1]-i[0]:.2f}' for i in spans_to_keep
 ]))
 EOF
 )"
@@ -61,7 +110,7 @@ for span in $(
    ) 2>&1 \
    | grep -E 'silencedetect|Parsed_mpdecimate_0|Duration: *[0-9]' \
    | grep -E 'silencedetect| (keep|drop) |Duration: *[0-9]' \
-   | python3 -c "$py" >&2
+   | python3 -c "$py"
 ); do
    n=$((n+1))
    ffmpeg -y -ss ${span%,*} -i "$f" -t ${span#*,} "$d/$(printf "%05d" $(ls "$d" | wc -l)).${f##*.}"