Instructions to use LiconStudio/Ltx2.3-VBVR-lora-I2V with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use LiconStudio/Ltx2.3-VBVR-lora-I2V 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", dtype=torch.bfloat16, device_map="cuda") pipe.load_lora_weights("LiconStudio/Ltx2.3-VBVR-lora-I2V") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
- Local Apps
- Draw Things
Great work, thank you!
Recently I was playing with this lora and it's awesome! As we all know ltx is quite step back to "bad hands" era, but with this lora in many cases it fixes the issue for me. Also in many I2V examples there's color shift at first few frames, seems this lora also improves that. Good work, can't wait for 240K and 480K release π
Recently I was playing with this lora and it's awesome! As we all know ltx is quite step back to "bad hands" era, but with this lora in many cases it fixes the issue for me. Also in many I2V examples there's color shift at first few frames, seems this lora also improves that. Good work, can't wait for 240K and 480K release π
Thankyou, but haven't you tested the new released official VBVR lora? Isn't it better?
Recently I was playing with this lora and it's awesome! As we all know ltx is quite step back to "bad hands" era, but with this lora in many cases it fixes the issue for me. Also in many I2V examples there's color shift at first few frames, seems this lora also improves that. Good work, can't wait for 240K and 480K release π
Thankyou, but haven't you tested the new released official VBVR lora? Isn't it better?
I didn't. You're talking about this one, right?
Recently I was playing with this lora and it's awesome! As we all know ltx is quite step back to "bad hands" era, but with this lora in many cases it fixes the issue for me. Also in many I2V examples there's color shift at first few frames, seems this lora also improves that. Good work, can't wait for 240K and 480K release π
Thankyou, but haven't you tested the new released official VBVR lora? Isn't it better?
I didn't. You're talking about this one, right?
Correct, due to the training process, I don't have extra GPU to test. So, I'm expecting a detail testing and feedback if you would like to.
Recently I was playing with this lora and it's awesome! As we all know ltx is quite step back to "bad hands" era, but with this lora in many cases it fixes the issue for me. Also in many I2V examples there's color shift at first few frames, seems this lora also improves that. Good work, can't wait for 240K and 480K release π
Thankyou, but haven't you tested the new released official VBVR lora? Isn't it better?
I didn't. You're talking about this one, right?
Correct, due to the training process, I don't have extra GPU to test. So, I'm expecting a detail testing and feedback if you would like to.
I did a quick few tests focused on anatomy. Official lora produces fewer anatomy issues (like 6 fingers) but it's not perfect, sometimes it struggles too. In one of my generations there was a text on a shirt while official lora was fine, yours managed to take this text from the shirt and rendered it as subtitles, then continuing the video other random subtitles showed on the video. Hope this feedback helps you somehow, for sure just a few tests is not good enough to be statistically significant. Color shift (few first frames) in both loras is a way less noticeable in my opinion compared to pure ltx 2.3
Recently I was playing with this lora and it's awesome! As we all know ltx is quite step back to "bad hands" era, but with this lora in many cases it fixes the issue for me. Also in many I2V examples there's color shift at first few frames, seems this lora also improves that. Good work, can't wait for 240K and 480K release π
Thankyou, but haven't you tested the new released official VBVR lora? Isn't it better?
I didn't. You're talking about this one, right?
Correct, due to the training process, I don't have extra GPU to test. So, I'm expecting a detail testing and feedback if you would like to.
I did a quick few tests focused on anatomy. Official lora produces fewer anatomy issues (like 6 fingers) but it's not perfect, sometimes it struggles too. In one of my generations there was a text on a shirt while official lora was fine, yours managed to take this text from the shirt and rendered it as subtitles, then continuing the video other random subtitles showed on the video. Hope this feedback helps you somehow, for sure just a few tests is not good enough to be statistically significant. Color shift (few first frames) in both loras is a way less noticeable in my opinion compared to pure ltx 2.3
Thank you for your feedback, I'll see if next version is better.