Mann-E Art

This is the new Stable Diffusion XL based model from our company Mann-E with more focus on artistic results closer to midjourney.

This model has been trained on around 1000 midjourney outputs in addition to photos gathered by our team. It is good in making photos, art, digital paintings, but may not be that great on pixel art or vector illustration.

How to use

For now, it is a diffusers model. We recommend using this piece of code:

from diffusers import DiffusionPipeline, DPMSolverSinglestepScheduler
import torch

pipe = DiffusionPipeline.from_pretrained(
    "mann-e/Mann-E_Art", torch_dtype=torch.float16
).to("cuda")

#This is equivalent to DPM++ SDE Karras, as noted in https://huggingface.co/docs/diffusers/main/en/api/schedulers/overview
pipe.scheduler = DPMSolverSinglestepScheduler.from_config(pipe.scheduler.config, use_karras_sigmas=True)

image = pipe(
  prompt="a cat in a bustling middle eastern city",
  num_inference_steps=8,
  guidance_scale=4,
  width=768,
  height=768,
  clip_skip=1
).images[0]
image.save("a_cat.png")

Additional Notes

SDXL 1.0 LoRas are working just fine with the model.
ControlNet, IPAdapter, InstantID are just fine.
Although we still recommend 768x768, but it is just working fine with bigger resolutions up to 1024x1024.
For rectangular images, 832x1216 and 608x1080 are working fine as well.
The model isn't good at vector or pixel without additional LoRA adapters.

Donations

Tron/USDT (TRC20) : TPXpiWACUZXtUszDbpLeDYR75NQTwngD8o
ETH (ERC20): 0x44e262f121b88bcb21caac3d353edd78c3717e08