Save Video+

Save Video+ is the save node you wish the base install shipped with. Where ComfyUI's built-in video saver gives you a filename and not much else, this one gives you a real codec choice (H.264 or H.265), a CRF quality dial, chroma subsampling control, audio muxing, workflow metadata embedding, and - the feature that makes it a workflow tool rather than a write-file node - an option to save the latent alongside the video so you can re-enter the pipeline losslessly.

It's the flagship of ComfyUI-FBnodes, François Beaudry's utility pack, and the rest of the pack is largely built around it: VACE Stitcher and Get Video Components+ both understand the .latent files it writes.

How it works

The video (VIDEO) input is what you're saving. Then the choices that matter:

• codec - h264 (8-bit, max compatibility) or h265/HEVC (10-bit, better compression and smoother gradients). If you're delivering somewhere with picky players, h264; if it's your own archive, h265.
• crf (0–51, default 18) - the constant-rate-factor quality dial. Lower is better; 0 is lossless, 18–23 is the sweet spot for high quality, 28+ gets small and crunchy. This is the single most useful control on the node, and the README's recommended range of 18–28 is a fair baseline.
• chroma - yuv420 (most compatible), yuv422 (better color for editing), or yuv444 (no chroma loss). Pick yuv444 and the node also generates a browser-compatible preview clip so you can still watch it in the UI - a nice touch since yuv444 files often won't preview directly.
• save (BOOLEAN) - OFF flips it into preview-only mode: fast encoding to temp, no permanent file. That's the "is this take worth keeping?" mode.
• save_latent (BOOLEAN) - writes a .latent file beside the video so Load Latent File / Get Video Components+ can recover the exact generation state. This is the pack's signature move.

The filename_prefix input supports date patterns like %date:yy-MM-dd% and %time% so you can organize output into dated folders without a filename-math node. Audio comes along via the AUDIO carried inside the VIDEO stream, workflow metadata is embedded in the file, and the node outputs the filename (STRING) it actually wrote - handy for wiring into a preview node or a notes field.

The inputs that matter

• video (VIDEO) - what you're saving (often the output of VACE Stitcher or Clip Stitcher).
• codec, crf, chroma - the quality stack.
• save - the preview-only toggle.
• save_latent - keep the lossless intermediate.
• latent (LATENT, optional) - the latent to save alongside, when it isn't already in the stream.

Installing it

ComfyUI Manager - search "ComfyUI-FBnodes" - or:

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

Restart ComfyUI. This node is the reason av (PyAV) is in requirements.txt - it does all the encoding. No model files.

Common issues

Two things trip people up. First, H.265 10-bit files frequently won't preview in a browser - that's what browser_compat and the auto-generated H.264 preview are for, so keep it ON unless you know the player handles HEVC. Second, crf is a quality target, not a size target: two clips at the same CRF can come out very different sizes depending on how noisy the content is - lower CRF for animation-style flat content, higher for noisy video, and expect to iterate. And if a saved video looks washed out or dark, that's almost always a VAE/decode issue upstream, not this node - it encodes what it's given.