Instructions to use WepeNerd/Obscura_Remova with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use WepeNerd/Obscura_Remova with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline from diffusers.utils import load_image, export_to_video # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("Lightricks/LTX-2.3", dtype=torch.bfloat16, device_map="cuda") pipe.load_lora_weights("WepeNerd/Obscura_Remova") prompt = "Remove the window curtains from the foreground." input_image = load_image("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/guitar-man.png") image = pipe(image=input_image, prompt=prompt).frames[0] export_to_video(output, "output.mp4") - Notebooks
- Google Colab
- Kaggle
- Local Apps
- Draw Things
Works with dev model?
Hi, i can see that you are using the dev_transformer_only_fp8_scaled version as your model. is the WF supposed to work as well with the 22b-dev model verison? if yes, the rest should stay the same?
Thanks!
So far on Dev model only works when Distilled LoRa is strength 1, not less. Best Distilled Lora for preserving original colors is 384-1. Obscura remova Lora should be between 1.3 and 1.6, maybe more. CFG should be 1. Apparently works better when feeding longer videos, maybe more than 48 frames long.
I really depends on the video input and what you're trying to change, but the distilled Lora strength can sometimes help if strength is higher, other times when strength is lowered, so it really is beneficial to experiment with those settings. And more frames is better (more information for it to reconstruct the scene).