Loop Forge

Tree full of falling leaves

Limit, from LTX 2.5 Creative Lab LoRAs, tested locally

Frame from the Still result
Still
Frame from the Cinemagraph LoRA, 1.5 result
Cinemagraph LoRA, 1.5
Watch it in the video at 4:45
LoRA
Cinemagraph at 1.5
Output
768×1056, 121 frames at 24fps
Sampling
8 steps, euler_ancestral
Seed
42
Generated in
6 min
Peak VRAM
11.6 GiB
Ran on
RTX 3060, 12GB (local)

What was wired in

Limit

A tree has thousands of small moving parts, and the whole frame drifts instead of only the leaves moving. I rewrote the prompt in the card's order and asked for just three falling leaves, with every leaf on the tree frozen, and it changed nothing.

CINEMAGRAPH_MOTION, tripod locked-off static camera, zero camera movement, only a few orange leaves are drifting down from the maple and the leaf tips are trembling in a light breeze, while the trunk, the bench, the grass and the sky stay perfectly still, seamless natural loop, faint birdsong.

Notes

The LoRA was trained at 512×704 with 25 frames. The two people above held far better at 121 frames than this did.

The Cinemagraph LoRA on Hugging Face

Run it yourself

Download everything for this test (.zip, 0.6 MB). It holds the workflow, every input it loads and a README. Copy the inputs into ComfyUI/input, put the models below in place and drag the workflow onto ComfyUI.

FilePut it inSize
ltx-2.5-22b-distilled-transformer-comfy-int8-convrot.safetensors gatedmodels/diffusion_models21.5 GB
gemma4-12b-with-proj-ltx-2.5-comfy-int8-convrot.safetensors gatedmodels/text_encoders15.4 GB
ltx-2.5-video-vae-bf16.safetensors gatedmodels/vae1.5 GB
ltx-2.5-audio-vae-bf16.safetensors gatedmodels/vae364.9 MB
ltx-2.5-22b-lora-cinemagraph-0.9.safetensors gatedmodels/loras201.5 MB

The Lightricks repositories (the LTX 2.5 base model and the LoRAs) are gated. Log in to Hugging Face and accept the terms on each repository page before downloading.

No custom nodes. Every node is part of ComfyUI 0.37.0 (commit 0f74f7fb, frontend 1.53.6).