🎬Video Output (Gradio)

Hua_Video_Output (🎬Video Output (Gradio)) is the video counterpart to the pack's image output node: it takes an IMAGE tensor of frames, encodes them into a video or animated image, and hands the result to the Gradio frontend the same way - write the file, drop the path in a temp JSON the webui polls, show the result in the video player. If you're packaging an AnimateDiff or video-gen workflow for someone, this is the node that finishes it.

It's basically a stripped-down VideoHelperSuite export: GIF/WebP go through imageio, everything else (mp4, webm, mp4-hevc, avi, mkv) goes through an ffmpeg subprocess that streams raw frames in on stdin. Frame rate, format, CRF quality, and encoding preset are all inputs. Optional extras include loop_count and pingpong (play forward then backward - a nice touch for loops) for GIF/WebP, an audio input for muxing sound, and a save_output flag that decides between the output dir and temp.

Inputs that matter for a beginner:

• images - the frames, i.e. your video model's output.
• format - pick from image/gif, image/webp, video/mp4, video/webm, video/mp4-hevc, video/avi, video/mkv. Default video/mp4.
• frame_rate - FLOAT, default 24.
• crf / preset - quality/speed tradeoff for the codec. Lower CRF = better quality, bigger file; default 23 is a sane middle.
• unique_id, name - the usual pack plumbing labels.

No outputs. Like the image node, the results show up in the webui's gallery/video player, not as a ComfyUI preview card.

Install: ComfyUI Manager → "ComfyUI_to_webui", or

cd ComfyUI/custom_nodes
git clone https://github.com/kungful/ComfyUI_to_webui.git

then restart. It needs ffmpeg for the video formats - the pack's requirements.txt includes imageio[ffmpeg], which bundles one, and the code falls back to an FFMPEG_PATH env var or a system install if you'd rather manage your own.

Where people get burned, three places. First: video input isn't in this pack. The README is explicit - if your workflow takes a video in, you need VideoHelperSuite's "Load Video (Upload)" node installed separately, and the frontend only shows the upload box when it finds that node. Second: the audio-muxing code currently has its ffmpeg audio path disabled in the shipped source ("diagnostic: audio processing temporarily disabled" is literally in the code), so an audio input is accepted but often silently ignored - don't build a workflow that depends on it. Third: it only saves video frames; if you're packaging a workflow that produces frames plus separate audio from a different node, this won't merge them today. For clean text-to-video or frame-to-gif packaging, though, it works and the pingpong option alone is worth having.