I2V Loop High/Low

What it is

The High/Low sibling of I2V Loop, and like that node it doesn't generate a thing. It's a config card that describes one segment for the pack's I2V Looper High/Low: a prompt, a frame count, and - the twist - two LoRA slots, one for the high-noise pass and one for the low-noise pass of Wan 2.2's two-expert sampling.

Why two LoRA slots is the whole point

Wan 2.2 splits denoising between a high-noise expert (motion, composition, camera) and a low-noise expert (detail). The community's standard optimization is to leave the high-noise pass untouched and put a speed LoRA on the low-noise pass only - you keep the scene composition and motion quality while cutting time on the detail pass. That's exactly what this node's lora_high / lora_low split is for. On a 2.2 setup it's a genuine reason to prefer this card over the plain I2VLoop.

Inputs

• prompt - the segment's text, multiline. Empty falls back to the base positive_prompt on the Looper, same behavior as the plain card.
• frames - default 81, range 1–200. Note there's no "0 = use 81" shortcut here like the plain I2VLoop has; the generator just clamps anything non-positive back to 81, so set the number you actually want.
• lora_high / lora_high_strength - LoRA for the high-noise pass, off by default, strength 0.8.
• lora_low / lora_low_strength - LoRA for the low-noise pass, the usual home for a speed LoRA.

Wiring

Send loop_config (type I2V_LOOP_HL) into loop_1 through loop_10 on I2V Looper High/Low, one card per segment, with unconnected slots skipped automatically. Don't plug this into the plain I2V Looper - the input types differ (I2V_LOOP_HL vs NATIVE_LOOP) and ComfyUI simply won't connect them.

Install

It ships in Native Looper with the rest of the pack:

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

Restart, or use Manager and search "Native Looper." No extra Python dependencies - only ComfyUI built-ins - but you still need the Wan 2.2 High/Low model pair, VAE, umt5 text encoder, and a CLIP Vision model before the Looper has anything to do with these cards. And a fair warning shared across the pack: it's a single-commit newborn with no community footprint yet, and the README even has a placeholder YOUR_USERNAME clone URL, so install from the real repo address above. Thin, focused, MIT-licensed - fine to build on, just don't expect a support forum.