Loop Forge

Running through Tokyo rain

Works, from LTX 2.5 Creative Lab LoRAs, tested locally

Frame from the Start still result
Start still
Frame from the No LoRA, slow motion in the prompt result
No LoRA, slow motion in the prompt
Frame from the Slow Motion LoRA, speed 0.2 result
Slow Motion LoRA, speed 0.2
Watch it in the video at 2:36
LoRA
Slow Motion Control at 1.0
Output
1920×1088, 121 frames at 24fps
Sampling
8 + 3 steps, euler_ancestral
Seed
42
Generated in
5 min
Peak VRAM
11.6 GiB
No LoRA run
Same graph without the LoRA, seed 42, 6 min
Ran on
RTX 3060, 12GB (local)

What was wired in

Works

The LoRA clip uses a plain caption of a normal-speed run, as the model card asks, and speed 0.2 alone slows it down. The clip without the LoRA has slow motion written into the prompt instead. That one runs faster and the effect comes and goes, while the LoRA holds it steady through the shot.

The prompts

Download workflowDownload everything for this test (.zip, 2.9 MB)
A medium close-up at night follows a young woman with long, curly copper-red hair and freckles as she runs down the middle of a narrow, wet Tokyo side street toward the camera, wearing a cream t-shirt and a bronze cross pendant on a long cord. Her arms pump, her curls bounce and swing around her face with each stride, and the pendant swings against her t-shirt. Light rain falls through the street, landing on her face and hair. The camera tracks her face, keeping it centred and in focus while it moves in a smooth arc around her, and the glowing neon signs, red paper lanterns and lit shop fronts behind her slide past, their pink, blue and amber light reflecting on the wet asphalt. The soundscape is her footsteps slapping on the wet street, her breathing, the patter of rain and the distant hum of the city, rendered with crisp high-resolution detail and moody, neon-lit cinematic colour.

Notes

Speed is set through the conditioning frame rate, which is 24 divided by the speed, so 120 here. Playback stays at 24fps. There is no trigger word.

Download the workflow without the LoRA (json)

The Slow Motion Control LoRA on Hugging Face

Run it yourself

Download everything for this test (.zip, 2.9 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-latent-spatial-upscaler-x2-bf16-1.0.safetensors gatedmodels/latent_upscale_models995.8 MB
ltx-2.5-22b-lora-slow-motion-control-1.0.safetensors gatedmodels/loras654.5 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).