Nodes/OmniNodes/Video Save 🎬
ComfyUI Node

Video Save 🎬

The last node in every video workflow

By TensorVizionΒ·Created 3 months agoΒ·Updated about 8 hours agoΒ· 0
Video Save 🎬
  • images
  • saved_path
  • summary
β—„fps24.0β–Ί
β—„filenametv_videoβ–Ί
β—„output_diroutput/tensorvizion/videoβ–Ί
β—„formatmp4β–Ί
β—„quality8β–Ί

Everything in a video workflow is a batch of images in memory right up until the end - and then you need it to be a file you can post, share, or import somewhere. The Video Save node from TensorVizion/OmniNodes is that final step: it encodes an IMAGE batch to MP4, WEBM, or GIF at a chosen frame rate, using imageio. It's the exit point the pack was built around - the node that turns all the batch processing above it into an actual deliverable.

How it works

The encode path depends on format. MP4 is written with the libx264 codec, WEBM with libvpx-vp9, and GIF via imageio.mimsave. quality (1–10, default 8) feeds the encoder. One detail worth knowing: filenames are collision-avoiding. If tv_video.mp4 already exists, it writes tv_video_001.mp4, then _002, and so on - repeat runs never silently overwrite a previous save. Same convention the pack uses for its other savers.

Inputs and outputs

Inputs: images (the frame batch), fps (1–120, default 24 - set this to your source FPS or the interpolation output rate), filename (default tv_video), output_dir (default output/tensorvizion/video), format (mp4/webm/gif), and quality. Outputs: saved_path (the exact file written) and summary (format, frame count, FPS, duration, path). It's an output node, so ComfyUI always runs it and won't cache past it.

The dependency you'll hit first

This is the one place in the pack where you will need a pip install, because imageio isn't bundled with ComfyUI. GIF works with plain imageio; MP4 and WEBM additionally need the ffmpeg plugin. If imageio is missing, the node doesn't crash - it returns a clear error in the summary telling you exactly what to install.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/TensorVizion/OmniNodes
pip install imageio imageio-ffmpeg

Install into ComfyUI's Python environment, restart. Under TensorVizion/Video.

Troubleshooting

  • "imageio is not installed" in summary - run the pip install above into the same Python that runs ComfyUI (not your shell's default Python).
  • MP4/WEBM fails but GIF works - the ffmpeg plugin is missing; pip install imageio-ffmpeg. The summary's error message says exactly this.
  • No file appears in ComfyUI's output tab - the default output_dir is output/tensorvizion/video relative to your ComfyUI root, and saved_path tells you the real absolute location. It may not show up in the browser's usual output list.
  • Wrong playback speed - check fps against how the frames were produced; a mismatch between generation FPS and save FPS is the classic cause of too-fast or too-slow clips.

Reach for it at the end of every video run. It's the boring node that makes the whole pipeline real.

CategoryTensorVizion/Video

Inputs (6)

NameTypeDefaultDescription
imagesIMAGEβ€”
fpsFLOAT24.01–120β€”
filenameSTRINGtv_videoβ€”
output_dirSTRINGoutput/tensorvizion/videoβ€”
formatCOMBOmp43 options: mp4, webm, gif
qualityINT81–10β€”

Outputs (2)

NameTypeDescription
saved_pathSTRINGβ€”
summarySTRINGβ€”