Image-to-Image
Diffusers
lora
template:diffusion-lora
text-to-video
image-to-video
video-to-video
lightx2v
Instructions to use rzgar/Bernini-R-LightX2V-4step-loras with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use rzgar/Bernini-R-LightX2V-4step-loras with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline from diffusers.utils import load_image # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("ByteDance/Bernini-R", dtype=torch.bfloat16, device_map="cuda") pipe.load_lora_weights("rzgar/Bernini-R-LightX2V-4step-loras") prompt = "Turn this cat into a dog" input_image = load_image("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/cat.png") image = pipe(image=input_image, prompt=prompt).images[0] - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- Draw Things
4+4 or 2+2
#5
by ZKong - opened
I saw your workflow use 4+4, but in my test 2+2 is good enough. very interesting。
yes bernini is a bit of a different beast and 4step is more like a safe zone.
for prompts that require drastic change to the scene or needs other non-existing characters into the scene, 2step output may be noisy, blurry, or have broken motion.
i2v is more forgiving, especially when the prompt is mainly about things that are already in the scene even regular lightx2v can work well with 2step,
the scene mostly gets blurry when something or someone appears or when the motion needs to be fast.
thanks
ZKong changed discussion status to closed