Instructions to use RuneXX/LTX-2.3-2.5-Workflows with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- LTX.io
How to use RuneXX/LTX-2.3-2.5-Workflows with LTX.io:
# Install the LTX-2 pipelines git clone https://github.com/Lightricks/LTX-2.git cd LTX-2 uv sync --frozen
# Download the weights from this repo, plus the Gemma text encoder hf download RuneXX/LTX-2.3-2.5-Workflows --local-dir models/LTX-2.3-2.5-Workflows hf download google/gemma-3-12b-it-qat-q4_0-unquantized --local-dir models/gemma-3-12b
# Fast pipeline (distilled model, no distilled LoRA needed) uv run python -m ltx_pipelines.distilled \ --distilled-checkpoint-path models/LTX-2.3-2.5-Workflows/<distilled-checkpoint>.safetensors \ --spatial-upsampler-path models/LTX-2.3-2.5-Workflows/<spatial-upsampler>.safetensors \ --gemma-root models/gemma-3-12b \ --prompt "A beautiful sunset over the ocean" \ --output-path output.mp4 # For image-to-video, add: --image path/to/image.jpg 0 0.8# HQ pipeline (two-stage, higher quality) uv run python -m ltx_pipelines.ti2vid_two_stages_hq \ --checkpoint-path models/LTX-2.3-2.5-Workflows/<checkpoint>.safetensors \ --distilled-lora models/LTX-2.3-2.5-Workflows/<distilled-lora>.safetensors 0.8 \ --spatial-upsampler-path models/LTX-2.3-2.5-Workflows/<spatial-upsampler>.safetensors \ --gemma-root models/gemma-3-12b \ --prompt "A beautiful sunset over the ocean" \ --output-path output.mp4 # For image-to-video, add: --image path/to/image.jpg 0 0.8 - Notebooks
- Google Colab
- Kaggle
Is there an equivalent workflow to WANGP 'Use LTX-2 raw format'?
Hi,
Thank you for the very useful workflows.
I have seen this wangp video and the results are very good. He is using 'Use LTX-2 raw format' rather than 'Transfer Human Motion'. I think Transfer Human Motion is using something like dwpose control net, but what's the equivalent of 'Use LTX-2 raw format'?
https://www.youtube.com/watch?v=VU_Jp5WCG7M
Thanks you for advance.
yes he is right. I hesitated to do that in the workflow. Because it almost seems like a "mistake" that it works.
Basically what it is, is not using pose estimation (those skeleton thingy from DW pose) at all .. but rather feed the raw input video as it is, straight into the model (no control net).
For some situations it works better. For dance, hands etc.
I'll update the workflow with a simple "Raw Mode" toggle ;-) (+ adding a variant that uses Realis Dance pose that Kijai made a lora for , that in some ways can work better than Ic-Lora-Union)