IAMCCS Video To Woosh Inputs

Foley - the sound-design pass where footsteps, cloth rustle, and ambient whooshes get added to silent footage - has a format problem. Your generated video is one thing, but the foley-generation side of the IAMCCS audio pipeline wants a specific downscaled, duration-capped video plus a plain frame batch. Video To Woosh Inputs is the adapter: it takes your VIDEO, thins and resizes it to fit the foley generator's constraints, and hands back both the image batch and the Woosh-ready video. The docstring calls it plainly: "Convert a generated Comfy VIDEO to IMAGE frames and WOOSH_VIDEO for foley generation."

How it works. Three knobs set the guardrails. max_duration_s (default 15) caps how much of the video you feed the foley pass - foley generation gets slower and noisier the longer the input, so a cap keeps it tractable (the pack's foley planner chunks into ~8-second windows). max_frames (default 144) caps the frame count by thinning the sequence if the source has more. max_side (default 384) caps the longest edge, resizing down so the foley generator isn't chewing on full-res frames it doesn't need. The output frame_rate is recomputed from what's left, so downstream timing stays honest.

The outputs. image_batch (the frames as an IMAGE batch, for anything that wants frames), woosh_video (the WOOSH_VIDEO type the pack's foley nodes consume), frame_rate (FLOAT - the effective rate after thinning, feed this forward rather than assuming 24), and report (what got trimmed/thinned/resized).

Where it fits. After your video generation/save node and before the foley generator - the pack's IAMCCS_CineWooshFoleyChunkPlanner is its natural downstream sibling, and the woosh_video output plugs straight into the Woosh side of the audio pipeline. If you're doing the LTX-2.3 "generate video, add sound in post" pattern, this is the bridge node.

Install. Pack-wide: ComfyUI Manager (search "IAMCCS"), or git clone https://github.com/IAMCCS/IAMCCS-nodes into custom_nodes, restart. It needs ComfyUI's VIDEO type to exist (recent ComfyUI with native video support); no extra pip packages.

Gotchas. The constraints are hard caps, not suggestions - a 40-second 1080p clip gets cut to 15 seconds and downscaled to 384 on the long edge, and that's the point, but it means the foley pass only ever sees a thumbnail-length version of your footage. Don't be surprised that foley synced to a heavily-thinned video drifts slightly at the tail; that's the cost of the cap. Also mind max_frames: 144 frames at 15 seconds is ~9.6 fps effective, which is fine for foley but wrong if you feed image_batch into something expecting real frame rates - that's exactly why the node recomputes frame_rate for you. Use it.