facefusion/facefusion · warning

restoring_audio_skipped

Error message

restoring_audio_skipped

What it means

This warning is logged by FaceFusion's video workflow when the standard audio-restoration branch fails: ffmpeg.restore_audio() (which copies audio from the target video into the temp processed video within trim_frame_start..trim_frame_end) returned False. As with the replace-audio branch it is non-fatal — the video pool is cleared and move_temp_file promotes the temp file to output_path — so the workflow completes but the output video is missing the original audio. It usually stems from FFmpeg being unable to extract/remux the audio for the requested frame range, an unsupported audio codec, or missing temp/output paths.

Source

Thrown at facefusion/workflows/to_video.py:207

		if source_audio_path:
			if ffmpeg.replace_audio(state_manager.get_item('target_path'), source_audio_path, state_manager.get_item('output_path')):
				video_manager.clear_video_pool()
				logger.debug(translator.get('replacing_audio_succeeded'), __name__)
			else:
				video_manager.clear_video_pool()
				if is_process_stopping():
					return 4
				logger.warn(translator.get('replacing_audio_skipped'), __name__)
				move_temp_file(state_manager.get_item('target_path'), state_manager.get_item('output_path'))
		else:
			if ffmpeg.restore_audio(state_manager.get_item('target_path'), state_manager.get_item('output_path'), trim_frame_start, trim_frame_end):
				video_manager.clear_video_pool()
				logger.debug(translator.get('restoring_audio_succeeded'), __name__)
			else:
				video_manager.clear_video_pool()
				if is_process_stopping():
					return 4
				logger.warn(translator.get('restoring_audio_skipped'), __name__)
				move_temp_file(state_manager.get_item('target_path'), state_manager.get_item('output_path'))
	return 0


def finalize_video(start_time : float) -> ErrorCode:
	if is_video(state_manager.get_item('output_path')):
		logger.info(translator.get('processing_video_succeeded').format(seconds = calculate_end_time(start_time)), __name__)
	else:
		logger.error(translator.get('processing_video_failed'), __name__)
		return 1
	return 0

View on GitHub (pinned to 4b1dedb853)

Solutions

  1. Verify FFmpeg presence/version (ffmpeg -version) and that the target's audio codec is supported (ffprobe target.mp4).
  2. Check that trim_frame_start/trim_frame_end are within the video's frame range and that the range contains audio.
  3. Ensure output_path directory exists and is writable, with enough disk space for the remux.
  4. Avoid interrupting the job during finalization so the is_process_stopping() branch isn't taken.
  5. If audio is still dropped, mux it back manually: ffmpeg -i out.mp4 -itsoffset <offset> -i target.mp4 -map 0:v -map 1:a -c copy final.mp4.

Example fix

# before
python facefusion.py run -s src.png -t clip.mp4 -o out.mp4 --trim-frame-start 0 --trim-frame-end 99999

# after (valid range within video; verified audio codec)
ffprobe -v error -show_entries stream=nb_frames,codec_name clip.mp4
python facefusion.py run -s src.png -t clip.mp4 -o out.mp4 --trim-frame-start 0 --trim-frame-end 250
Defensive patterns

Strategy: fallback

Validate before calling

import subprocess, json, shutil

def audio_restorable(path: str, frame_start: int, frame_end: int) -> bool:
    if not shutil.which('ffprobe'):
        return False
    probe = subprocess.run(['ffprobe', '-v', 'error', '-show_entries', 'stream=nb_frames,codec_type,codec_name', '-of', 'json', path], capture_output=True, text=True)
    if probe.returncode != 0:
        return False
    streams = json.loads(probe.stdout).get('streams', [])
    has_audio = any(s.get('codec_type') == 'audio' for s in streams)
    nb_frames = max(int(s.get('nb_frames', 0) or 0) for s in streams if s.get('codec_type') == 'video') or None
    in_range = nb_frames is None or (0 <= frame_start < nb_frames and frame_end >= frame_start)
    return has_audio and in_range

When it happens

Trigger: Running a video job without the conditional-audio option where ffmpeg.restore_audio(target_path, output_path, trim_frame_start, trim_frame_end) returns False because: (1) trim_frame_start/trim_frame_end define a range with no matching audio, (2) the target's audio codec is unsupported by the installed FFmpeg, (3) the temp processed video was removed (video pool cleared) or output_path is unwritable, (4) the process is stopping (the adjacent is_process_stopping() check returns exit code 4 before the warn in that case), or (5) FFmpeg is absent/outdated.

Common situations: Using --trim-frame-start/--trim-frame-end values that clip outside the audio timeline; processing videos with exotic audio (e.g. AC-3, Opus in odd containers) on a slim FFmpeg build; CI containers without FFmpeg encoders; cancelled jobs; outputs ending up silent and users filing 'no audio in result' issues.

Related errors

AI-assisted analysis of facefusion/facefusion@4b1dedb853 (2026-08-28). Data as JSON: /api/errors/efce38405ddf6e2f. Report an issue: GitHub.