IAMCCS NextFrame Commit Preview

Storyboard work has a chicken-and-egg problem: you generate a frame for the next scene, and then you need that exact image to be the input for the scene after that. IAMCCS_NextFrameCommitPreview is the node that closes the loop - it takes whatever frames you just generated, saves them as real PNG files into ComfyUI's input folder, and appends them to your storyboard JSON so the next generation can use them as its source. Display name: NextFrame Commit Preview.

"Commit" is the operative word. Everything else in the NextFrame flow produces a transient image in memory; this node makes it durable. It writes each frame as a timestamped PNG into ComfyUI/input, naming it with your session and a running frame number (IAMCCS_NextFrame_<session>_NNN.png), bumps the counter past anything already in the storyboard, and tags the file with PNG metadata - the prompt, the seed, your workflow, the whole frame record - so the file itself carries its provenance. If you've loaded a source_image that isn't already in the storyboard, it registers it as the first "source frame" entry first. The storyboard JSON it returns is capped at the most recent 96 frames, which keeps the JSON from becoming a monster on very long sessions.

It's an output node, which means it's also where the UI side of the NextFrame workflow reports back: the frame count, the "Frame N ready" message, and the generated filename all flow through its UI output.

Inputs and outputs

Required inputs: images (the batch to save), source_image (optional string - the current scene's source file, registered into the board), prompt_text, storyboard_json (the running board), session_id, seed, and settings_json (extra settings recorded into each frame's metadata). Outputs: image (passes the frames through), storyboard_json (the updated board), and generated_filename (the file you'll feed into the next builder).

The honest take

The genius is in the filename output. IAMCCS_NextFrameBuilder takes a source image path as its primary input, and this node hands you exactly the path of the file it just committed - so the canonical loop is: build a frame, commit it, take the generated_filename, feed it back as the next source_image. The files land in your input folder, which is also where LoadImage-style nodes and the reference workflow expect them, so there's no tensor-copying or temp-preview indirection.

Two practical notes. First, the frame numbers come from the storyboard JSON, not the filename - keep the storyboard connected or numbering will drift. Second, this node doesn't generate anything; it persists. If you wire raw noise into images, you'll commit raw noise with a clean conscience and a confusing filename. Commit the output of the actual generation chain.

Install

It's part of IAMCCS-nodes:

cd ComfyUI/custom_nodes
git clone https://github.com/IAMCCS/IAMCCS-nodes

or via ComfyUI Manager (search "IAMCCS"), then restart. It has no heavy model requirements of its own - it's pure bookkeeping plus PNG save - but it earns its keep inside the NextFrameBuilder storyboard flow, which does pull the Qwen Image Edit stack. If that's the plan, expect to download the Qwen Image Edit 2511 model, its CLIP and VAE, and the LoRAs the reference workflow names.