Instructions to use Lightricks/LTX-2.3-22b-IC-LoRA-HDR with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use Lightricks/LTX-2.3-22b-IC-LoRA-HDR with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("Lightricks/LTX-2.3-22b-IC-LoRA-HDR", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
running into this issue
not sure what it is tbh. alot of distortion and noise as well as overexposure even though its on 1.0. the reference video itself does have some noise and artifacting but the lora seems to amplify it
reference video
LTX output
Seems like are using a distilled lora in top of the distilled model. you should remove the distilled lora in that case
Omg 🤦♂️🤣🤣🤣🤣🤣 yoyr the best
un fortunately the issue still persists even when replacing the distilled lora with the correct scene embedding lora and pairing it with the hdr lora itself. (not sure if thats correct but claude said it is so i assume it is as well)
ill continue to adjust and see if i cant correct it. it may be the Q2 gguf format compression causing it? not sure
workflow REQUIRES dev model and distill lora, it will NOT work with the distilled model itself. problem solved, waiting for a generation to provide evidence!
well, running the dev model doesnt seem to fic the issue either... im not sure. ill continue to trouble shoot!
Its the kj vae loader, needed updated. Ill close the discussion!
correct scene embedding lora
hi, there, if you don't mind me asking what is scene embedding lora? what does it do? Thx!
correct scene embedding lora
hi, there, if you don't mind me asking what is scene embedding lora? what does it do? Thx!
not sure because its not needed in the workflow at all, i assumed it was a part of the sequence but i dont think it plays a role here. i didnt use it to get a good output
not sure because its not needed in the workflow at all, i assumed it was a part of the sequence but i dont think it plays a role here. i didnt use it to get a good output
I mean could you possibly give me a link to where you downloaded it from? I'm collecting info on LoRas for my little documentation project wanx-troopers.github.io and this is one I haven't heard about yet
its literally in this repo bud
