Lightricks

/

LTX-2.5-Diffusers

like 41

Follow

LTX.io 4.78k

This PR is a follow up to #13 that adds the prompt_enhancer and processor from google/gemma-4-E2B-it to the repo for the standard pipeline (which unlike the modular pipeline cannot be configured to directly load the weights from the Gemma 4 repo). Unlike #13, it doesn't accidentally duplicate the prompt_enhancer weights by including both model.safetensors and model shards.

You can test the changes by loading the pipeline with the refs/pr/15 revision corresponding to this PR:

import torch
from diffusers import LTX2Pipeline

pipe = LTX2Pipeline.from_pretrained("Lightricks/LTX-2.5-Diffusers", revision="refs/pr/15", torch_dtype=torch.bfloat16)
...