Save Video

Every generation in this pack ends at the same node, and it's the least glamorous part of the whole thing: TurboDiffusionSaveVideo takes the frame batch your sampler produces and writes it to disk as an MP4, GIF, or WebM. There's no cleverness to it, no hidden settings that unlock quality. It just needs to work, because the alternative is watching your frames flash past in a preview and realizing you forgot to save them.

It's an output node with no outputs. You feed it frames (whatever the TurboDiffusionI2VSampler gave you, or any image batch) and it writes a timestamped file - prefixed with your filename_prefix - into a turbodiffusion_videos/ subfolder of ComfyUI's output directory. That subfolder is a small gotcha: it's not in the root of output/, it's in ComfyUI/output/turbodiffusion_videos/.

The few settings that matter

• format - mp4, gif, or webm. MP4 is the one for anything you care about; GIF is for quick sharing on Discord/Reddit; WebM is the lightweight web-player option.
• fps - default 24. For typical Wan clips that's right; drop to 16 if you want the smoother old-school video look or smaller files.
• filename_prefix - default turbodiffusion. A timestamp gets appended automatically, so you won't overwrite previous runs. If you're chaining many clips, prefix them with something meaningful now, not later.
• quality (optional) - 1-10, default 8, only applies to MP4/WebM. 8 is a fine middle; drop to 6-7 for smaller files if you're iterating.
• optimize_gif (optional) - default on, GIF-only. Keep it on; it's the difference between a 40 MB meme and a 10 MB one. GIFs save looping forever by default, which is usually what you want for sharing.

The dependency you didn't expect

MP4/WebM encoding isn't bundled with ComfyUI, so the node needs opencv-python or imageio + imageio-ffmpeg. GIF export works through Pillow, which you already have. The pack's README install line (pip install einops loguru omegaconf pandas) doesn't mention any of this, and the code will tell you - in an unhelpful way - if the package is missing. Run this once and get ahead of it:

pip install imageio imageio-ffmpeg

If you'd rather have OpenCV do the MP4 work, pip install opencv-python covers it, though imageio is the preferred path in the code (cv2 is only a fallback). ComfyUI's own runtime sometimes pulls imageio in anyway, which is why the node "just works" on some installs and dies on others.

Troubleshooting

• ImportError about opencv or imageio - that's the missing dependency above. Install it, restart, done.
• File isn't where you expect - check ComfyUI/output/turbodiffusion_videos/, named {prefix}_{timestamp}.{ext}. It won't be in the pack's directory or the root of output/.
• GIF looks blocky - that's GIF being GIF, not a setting you missed. Export MP4 if you want the actual quality you generated.