🌊🎥 FlowState Video Creator

Modern ComfyUI video generation doesn't end at the sampler. Samplers hand you a batch of frames; you still need to bundle those frames into a video container and write it to disk. Core ComfyUI splits that across a CreateVideo node and a SaveVideo node, and FlowState Video Creator is just those two fused into one box, with the filename and format knobs on top. If you've got a frames batch and want an mp4 out the other side, this is the whole pipeline in one node.

How it works

Feed it an IMAGE batch (any frames output - from this pack's WAN Studio, from a video model's decode, from anywhere), pick an fps, and the node runs ComfyUI's own CreateVideo to assemble the frames into a video container, then SaveVideo to write it to ComfyUI/output. It's an output node - no sockets out, because its job is the file on disk. The format (mp4 / auto) and codec (h264 / auto) dropdowns default sensibly and mostly only matter if you need a specific container for another tool.

Two details are worth knowing:

• audio (AUDIO) is optional and the quiet killer feature - wire in an audio track and it gets muxed into the saved video. That's how you go from silent frames to a finished clip with a music bed or voiceover in one node instead of a post-processing trip to ffmpeg.
• filename_prefix accepts ComfyUI's formatting tags - %date:yyyy-MM-dd%, or values pulled from other nodes like %Empty Latent Image.width% - so you can build organized filenames for batch runs. Default is video/WANStudio, which tells you who this node was made for.

Where it fits

It's the terminal node of a FlowState video graph: WAN Studio or another video sampler produces the frames, this writes the file. It's also genuinely usable standalone - if you're building video graphs with other packs, dropping this on the end of any frames output gives you the create-plus-save combo and the audio muxing without hunting for the right core nodes. The tradeoff is the same as every combo node: it always saves. There's no "preview only" mode here, because that's what the sibling FlowState_VideoPreview node is for.

Install

Standard FlowState Creator Suite install:

cd ComfyUI/custom_nodes
git clone https://github.com/flowstateeng/FlowState-Creator-Nodes

then restart, or search FlowState-Creator-Nodes in ComfyUI Manager. The pack has no pip dependencies - requirements.txt is empty - and this node just wraps the video tooling ComfyUI already ships. No model files to download.

The gotcha to expect

Because it always saves, you'll find yourself accumulating files while iterating on a video - every run writes to output/video/. If you're dialing in a prompt and only want to look at the result, wire the frames into FlowState_VideoPreview instead and save your disk. And if your exported video plays jittery, it's almost always the fps you chose for the frames, not the codec - match it to what the model was trained for (Wan tends to like 12–16).