Last active
September 24, 2025 16:31
-
-
Save Interpause/4aa43ebc2b04eecec31bac1994f08f08 to your computer and use it in GitHub Desktop.
Snippet to change video speed w/o reencoding video while reencoding audio to maintain pitch
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| display(Markdown("### Preparing Videos")) | |
| if dataset_df is None: | |
| raise RuntimeError("Dataset DF unavailable.") | |
| all_video_ids = sorted(list(dataset_df['video_id'].dropna().unique())) | |
| logger.info(f"Processing {len(all_video_ids)} unique video IDs.") | |
| skipped = 0 | |
| processed = 0 | |
| vid_paths = list(extracted_videos_path.glob("*.mp4")) | |
| for vid_path in tqdm(vid_paths, desc="Transforming Videos", unit="video"): | |
| out_path = speed_videos_path / vid_path.name | |
| if out_path.is_file(): | |
| skipped += 1 | |
| continue | |
| vid_path = str(vid_path.resolve()) | |
| out_path = str(out_path.resolve()) | |
| # I assume all youtube videos are h264 encoded. | |
| tf_bitstream = tempfile.NamedTemporaryFile(delete=False, suffix=".h264") | |
| tf_audio = tempfile.NamedTemporaryFile(delete=False, suffix=".aac") | |
| tf_final = tempfile.NamedTemporaryFile(delete=False, suffix=".mp4") | |
| tf_bitstream.close() | |
| tf_audio.close() | |
| tf_final.close() | |
| # Get original FPS. | |
| result = subprocess.run( | |
| ["ffprobe", "-v", "error", "-select_streams", "v", "-of", "default=noprint_wrappers=1:nokey=1", "-show_entries", "stream=r_frame_rate", vid_path], | |
| check=True, | |
| capture_output=True, | |
| ) | |
| fps = float(fractions.Fraction(result.stdout.decode().strip())) | |
| new_fps = fps * VIDEO_SPEED_FACTOR | |
| # Extract the audio from the mp4 file while also slowing/speeding it up. | |
| factor = VIDEO_SPEED_FACTOR | |
| filter = [] | |
| while factor < 0.5: | |
| filter.append("atempo=0.5") | |
| factor /= 0.5 | |
| filter.append(f"atempo={factor:.6f}") | |
| filter = ",".join(filter) | |
| subprocess.run( | |
| ["ffmpeg", "-y", "-i", vid_path, "-vn", "-filter:a", filter, "-c:a", "aac", tf_audio.name], | |
| check=True, | |
| stderr=subprocess.DEVNULL, | |
| stdout=subprocess.DEVNULL, | |
| ) | |
| # Extract the h264 bitstream from the mp4 file. | |
| subprocess.run( | |
| ["ffmpeg", "-y", "-i", vid_path, "-map", "0:v", "-c:v", "copy", "-bsf:v", "h264_mp4toannexb", tf_bitstream.name], | |
| check=True, | |
| stderr=subprocess.DEVNULL, | |
| stdout=subprocess.DEVNULL, | |
| ) | |
| # Remux the bitstream with new audio. | |
| subprocess.run( | |
| ["ffmpeg", "-y", "-fflags", "+genpts", "-r", f"{new_fps:.6f}", "-i", tf_bitstream.name, "-i", tf_audio.name, "-c:v", "copy", "-c:a", "copy", tf_final.name], | |
| check=True, | |
| stderr=subprocess.DEVNULL, | |
| stdout=subprocess.DEVNULL, | |
| ) | |
| shutil.move(tf_final.name, out_path) | |
| processed += 1 | |
| Path(tf_bitstream.name).unlink(missing_ok=True) | |
| Path(tf_audio.name).unlink(missing_ok=True) | |
| break | |
| print(f"\n\n{skipped} videos skipped, {processed} videos processed, {len(vid_paths)} total.") |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment