renato97/viral_project
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: speed up output clips to 1.25x speed by default and scale subtitle/mute timestamps

Browse Source
This commit is contained in:
Renato
2026-07-02 19:38:50 +02:00
parent 05e0217541
commit 174b992335
2 changed files with 34 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files
openshorts/main.py
+27 -11
View File
@@ -887,6 +887,7 @@ if __name__ == '__main__':
parser.add_argument('-o', '--output', type=str, help="Output directory or file (if processing whole video).") parser.add_argument('-o', '--output', type=str, help="Output directory or file (if processing whole video).")
parser.add_argument('--keep-original', action='store_true', help="Keep the downloaded YouTube video.") parser.add_argument('--keep-original', action='store_true', help="Keep the downloaded YouTube video.")
parser.add_argument('--skip-analysis', action='store_true', help="Skip AI analysis and convert the whole video.") parser.add_argument('--skip-analysis', action='store_true', help="Skip AI analysis and convert the whole video.")
parser.add_argument('--speed', type=float, default=1.25, help="Speed up factor for the output clips (default: 1.25).")
args = parser.parse_args() args = parser.parse_args()
@@ -1020,16 +1021,30 @@ if __name__ == '__main__':
clip_temp_path = os.path.join(output_dir, f"temp_{clip_filename}") clip_temp_path = os.path.join(output_dir, f"temp_{clip_filename}")
clip_final_path = os.path.join(output_dir, clip_filename) clip_final_path = os.path.join(output_dir, clip_filename)
# ffmpeg cut # ffmpeg cut & speed up
cut_command = [ if args.speed != 1.0:
'ffmpeg', '-y', setpts = 1.0 / args.speed
'-ss', str(start), cut_command = [
'-to', str(end), 'ffmpeg', '-y',
'-i', input_video, '-ss', str(start),
'-c:v', 'libx264', '-crf', '18', '-preset', 'fast', '-to', str(end),
'-c:a', 'aac', '-i', input_video,
clip_temp_path '-filter_complex', f'[0:v]setpts={setpts:.6f}*PTS[v];[0:a]atempo={args.speed:.6f}[a]',
] '-map', '[v]', '-map', '[a]',
'-c:v', 'libx264', '-crf', '18', '-preset', 'fast',
'-c:a', 'aac',
clip_temp_path
]
else:
cut_command = [
'ffmpeg', '-y',
'-ss', str(start),
'-to', str(end),
'-i', input_video,
'-c:v', 'libx264', '-crf', '18', '-preset', 'fast',
'-c:a', 'aac',
clip_temp_path
]
subprocess.run(cut_command, stdout=subprocess.DEVNULL, stderr=subprocess.PIPE) subprocess.run(cut_command, stdout=subprocess.DEVNULL, stderr=subprocess.PIPE)
# Generate SRT for this clip and get mute ranges for profanity # Generate SRT for this clip and get mute ranges for profanity
@@ -1038,7 +1053,8 @@ if __name__ == '__main__':
success_srt, mute_ranges = generate_srt( success_srt, mute_ranges = generate_srt(
transcript, start, end, srt_path, transcript, start, end, srt_path,
max_chars=20, max_duration=2.0, max_chars=20, max_duration=2.0,
swears_path=swears_path, return_mute_ranges=True swears_path=swears_path, return_mute_ranges=True,
speed_factor=args.speed
) )
# Process vertical, passing the relative mute ranges for profanity muting # Process vertical, passing the relative mute ranges for profanity muting
openshorts/subtitles.py
+7 -7
View File
@@ -90,7 +90,7 @@ def _merge_overlapping(ranges, pad_seconds=0.05):
merged.append((s, e)) merged.append((s, e))
return merged return merged
def generate_srt(transcript, clip_start, clip_end, output_path, max_chars=20, max_duration=2.0, swears_path=None, return_mute_ranges=False): def generate_srt(transcript, clip_start, clip_end, output_path, max_chars=20, max_duration=2.0, swears_path=None, return_mute_ranges=False, speed_factor=1.0):
""" """
Generates an SRT file from the transcript for a specific time range. Generates an SRT file from the transcript for a specific time range.
Groups words into short lines suitable for vertical video. Groups words into short lines suitable for vertical video.
@@ -110,9 +110,9 @@ def generate_srt(transcript, clip_start, clip_end, output_path, max_chars=20, ma
cleaned_word = re.sub(r'[^\w]', '', word_text.lower()) cleaned_word = re.sub(r'[^\w]', '', word_text.lower())
if cleaned_word in swears: if cleaned_word in swears:
# Record absolute mute range relative to clip start # Record absolute mute range relative to clip start, scaled by speed_factor
rel_start = max(0.0, word_copy['start'] - clip_start) rel_start = (max(0.0, word_copy['start'] - clip_start)) / speed_factor
rel_end = max(0.0, word_copy['end'] - clip_start) rel_end = (max(0.0, word_copy['end'] - clip_start)) / speed_factor
raw_mute_ranges.append((rel_start, rel_end)) raw_mute_ranges.append((rel_start, rel_end))
# Censor word # Censor word
word_copy['word'] = censor_word(word_text) word_copy['word'] = censor_word(word_text)
@@ -134,9 +134,9 @@ def generate_srt(transcript, clip_start, clip_end, output_path, max_chars=20, ma
block_start = None block_start = None
for i, word in enumerate(words): for i, word in enumerate(words):
# Adjust times relative to clip # Adjust times relative to clip, scaled by speed_factor
start = max(0, word['start'] - clip_start) start = (max(0, word['start'] - clip_start)) / speed_factor
end = max(0, word['end'] - clip_start) end = (max(0, word['end'] - clip_start)) / speed_factor
# Word info copy with relative times # Word info copy with relative times
w_rel = { w_rel = {