Instructions to use Lightricks/LTX-2.5-22b-IC-LoRA-Refine-Details with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- LTX-2
How to use Lightricks/LTX-2.5-22b-IC-LoRA-Refine-Details with LTX-2:
# Install the LTX-2 pipelines git clone https://github.com/Lightricks/LTX-2.git cd LTX-2 uv sync --extra natten
# Download the adapter weights from this repo # (base components come from Lightricks/LTX-2.5 — see Files and versions) hf download Lightricks/LTX-2.5-22b-IC-LoRA-Refine-Details --local-dir models/LTX-2.5-22b-IC-LoRA-Refine-Details
# Video-to-video with the IC-LoRA (runs on the distilled LTX-2.5 base) uv run python -m ltx_pipelines.ic_lora \ --transformer-path path/to/distilled-transformer.safetensors \ --text-encoder-path path/to/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors \ --video-vae-path path/to/video-vae.safetensors \ --audio-vae-path path/to/audio-vae.safetensors \ --spatial-upsampler-path path/to/spatial-upsampler.safetensors \ --lora models/LTX-2.5-22b-IC-LoRA-Refine-Details/<weights>.safetensors 1.0 \ --video-conditioning reference.mp4 1.0 \ --prompt "your prompt here" \ --output-path output.mp4 - Notebooks
- Google Colab
- Kaggle
what kind of beast do you need to run this thing? 12gb vram not enough?
#3
by douchebanner - opened
i tried with a 512*384, 97 frames clip and i had to cancel after 40 minutes.
[INFO] Model LTXAV prepared for dynamic VRAM loading. 20484MB Staged. 480 patches attached. Force pre-loaded 608 weights: 3303 KB.
[INFO] Interrupting prompt d9afac6d-e229-487b-b25e-9267dfe03d2b
[INFO] Comfy model compiler graph breaks: 1, rogues: 1
[INFO] Processing interrupted
[INFO] Prompt executed in 00:41:59
forgot to mention i selected hd, full hd and changed the size to 512*384 in the Get Tiling Sizes node
Are you using comfy? make sure you use quantized models, including gemma4
did you also do chunking ?