Instructions to use Tongyi-MAI/Z-Image with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use Tongyi-MAI/Z-Image with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("Tongyi-MAI/Z-Image", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Inference
- Notebooks
- Google Colab
- Kaggle
- Local Apps
- Draw Things
- DiffusionBee
What is the best settings for this model? CGF, Sampler, Scheduler, Sampling Steps?
#12
by mrmaq01 - opened
What is the best settings for this model? CGF, Sampler, Scheduler, Sampling Steps?
res2s beta, or bong tangent looks great. I use cfg 4 or 4.5. shift at 3.
euler + simple/beta/sgm_uni, anything else will only get a twisted mess of Cthulhu.
I can see the potential with Z-Image Base. However, I don't think that KSampler is the optimal sampler to use. Isn't there a specialized version around? I have found one, but this seems to be designed for video, not pictures and claims that a dimension is missing - probably time..
I also wonder if this still is holding ground: https://myaiforce.com/z-image-samplers-schedulers/
according to the tests of this author, best settings for Z-Image base are
DPMPP SDE + Beta
DPMPP SDE + DDIM Uniform
Euler Ancestral + Beta
Euler Ancestral + DDIM Uniform