I2V Loop

What it actually is

First thing to get straight: I2V Loop doesn't generate anything. No model inputs, no video out, nothing to preview. It's a config node - the per-segment script for one act of a longer movie. You build a handful of these little cards, each carrying a prompt, a frame count, and an optional LoRA, then feed them into the pack's I2V Looper node, which does the actual Wan image-to-video generation.

This split is the whole point of the Native Looper design. ComfyUI's built-in Wan support gives you one clip per generation - usually 81 frames, a few seconds - and then the scene starts drifting if you push it. To get a long, continuous take you chain clips, and I2V Loop is how you tell the chain what each link should be. It's the boring half of the pack, but it's the half that makes long videos possible without doing everything by hand.

The inputs that matter

• prompt - the text driving this segment. Multiline, so don't be shy. Leave it empty and the Looper quietly falls back to its base positive_prompt, which is how you end up with duplicate scenes you didn't plan.
• frames - how long the segment runs. The gotcha is right in the tooltip: 0 = use 81. Wan's default clip length. If you set 0 expecting "no frames," you get a full 81-frame segment.
• lora_name / lora_strength - an optional per-segment LoRA, off by default at strength 0.8. Useful when different scenes in one video want different styles, though honestly most of my loops run with it left on "None."

Wiring it up

Feed loop_config into loop_1 (or loop_2 through loop_10) on I2V Looper - one I2VLoop node per segment. The Looper auto-detects how many slots are actually connected and skips the rest, so you can run a 2-segment loop or go all the way to ten. Just don't wire loop_config into a Preview or VHS node expecting pixels; you'll get nothing but confusion.

Installing the pack

I2V Loop ships inside Native Looper, so install the whole pack once:

cd ComfyUI/custom_nodes
git clone https://github.com/masteroleary/ComfyUI-LooperNode

Then restart ComfyUI. Or use ComfyUI Manager and search "Native Looper." There's no requirements.txt - the pack only calls ComfyUI built-ins (WanImageToVideo conditioning, KSampler, CLIP Vision encode), so there are no extra Python dependencies to fight. What you do need is a Wan I2V model sitting in your models/diffusion_models folder, because this node is useless without the Looper to drive it.

One honest warning: this pack is brand new - at the time of writing it's a single commit with zero community footprint, and the repo's own README still contains a placeholder YOUR_USERNAME clone URL. Trust the repository address above, not the README's copy-paste. For a node this thin that's fine; just know you're an early adopter and there's no support forum full of answers waiting for you yet.