Instructions to use rzgar/LTX-2.3-Motion-Enhancer-n4w with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use rzgar/LTX-2.3-Motion-Enhancer-n4w 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("rzgar/LTX-2.3-Motion-Enhancer-n4w") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Inference
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- Draw Things
scary and ugly scars on body
hight change we will get scary and ugly scars on body in upscale period. i wonder if it is the lora or ltx 's problem. I have not met this problem before
hight change we will get scary and ugly scars on body in upscale period. i wonder if it is the lora or ltx 's problem. I have not met this problem before
Can you share a video with the workflow embedded? It would make it easier to pinpoint the issue. Usually, if skin or deformation occurs, it’s mostly due to a heavily quantized text encoder or the base model being overpowered. *cfg_pp samplers also can be problematic sometimes in the base model is distill
only change the models, using the comfyui ltx2.3 i2v template.
appears on the body part hidden from the first frame, for example when hand moves, a scar shows in hand's last position . change a seed in upscale peroid may solve , but seems high chance.
I tried other lora to generate a few tests, none of them has scars
gemma in4
only change the models, using the comfyui ltx2.3 i2v template.
appears on the body part hidden from the first frame, for example when hand moves, a scar shows in hand's last position . change a seed in upscale peroid may solve , but seems high change.
I tried other lora to generate a few tests, none of them has scars
Gemma int4 is the issue. try gemma-3-12b-it-heretic-v2_fp8_e4m3fn.safetensors if your hardware allows, the lora itself is rank126 bf16
thanks
