🔥AddAudio

AddAudio does one simple thing: mux an audio file onto a video. You have a silent mp4 (every AI-generated video ever, plus anything that came out of Frames2Video without an audio_path) and you want music or voiceover on it. The nice part - and the reason you should reach for this over a generic converter - is that it stream-copies the video track, so the video itself is never re-encoded. That makes it effectively instant and lossless; the source video's quality is untouched.

How it works

It runs FFmpeg with -c:v copy -c:a copy, which means no re-encoding on either side: the video bytes are copied as-is and the audio stream is copied in. The only real processing is the delay_play offset, applied with -itsoffset so your audio can start a few seconds late - handy for syncing narration to something that happens partway through. One honest limitation: stream copy can't fix mismatched durations. If the video is longer than the audio, the tail is silent; if shorter, the audio gets cut at the video's end. Fine for music beds, fiddly for precisely-timed voiceover.

The inputs that matter

• video_path - the silent video. Must exist and have a video extension.
• audio_from - audio_file or video_file. The clever bit: you can pull the audio out of a second video instead of a standalone audio file, which saves you an ExtractAudio step when you want to transplant someone's soundtrack from one clip to another.
• file_path - where the audio (or second video) lives. If you pick video_file, the node verifies that file actually has an audio track and errors out if it doesn't - that's a real "I picked the wrong file" tripwire, and it's nice to have.
• delay_play - seconds of silence before the audio starts, default 0.
• output_path - pre-existing directory for the result (auto-named with a timestamp).

Outputs

A single video_complete_path string pointing at your new, now-audible file. Nothing else, no preview.

Install

This pack's install is the same everywhere, and it's refreshingly light: requirements.txt is just pathlib. The actual prerequisite is an FFmpeg binary on your PATH - every node here is a wrapper around the ffmpeg command. Install via ComfyUI Manager (search "ComfyUI-FFmpeg") or:

cd ComfyUI/custom_nodes
git clone https://github.com/MoonHugo/ComfyUI-FFmpeg
cd ComfyUI-FFmpeg
pip install -r requirements.txt

Then restart ComfyUI.

Gotchas

• If the video already has a soundtrack, adding another doesn't mix them - FFmpeg's -map 0:v -map 1:a here replaces the old audio with the new one. If you wanted them layered, that's not this node.
• Timestamped output names mean running the same video through twice gives you two files rather than overwriting. That's by design across the whole pack.
• No -shortest flag is used, so a long audio file gets truncated at the video length rather than extending the video - don't expect this node to stretch the picture to fit your audio.