Simple Diffusion XS -now live here https://huggingface.co/AiArtLab/sdxs-1b

Its dev repo!

This is a research repository for experimenting with different architectures

At AiArtLab, we strive to create a free, compact and fast model that can be trained on consumer graphics cards.

• Model: 1.5b parameters
• Text encoder: Qwen3 (0.6B parameters)
• VAE: 8x16ch, Simple VAE

The model was trained for ~3 months on (2-4)x RTX 5090 GPUs using over 1+ million images of various resolutions and styles, primarily anime and illustrations.

Gradio

Example

import torch
from diffusers import DiffusionPipeline

device = "cuda" if torch.cuda.is_available() else "cpu"
dtype = torch.float16 if torch.cuda.is_available() else torch.float32

pipe_id = "AiArtLab/sdxs"
pipe = DiffusionPipeline.from_pretrained(
    pipe_id,
    torch_dtype=dtype,
    trust_remote_code=True
).to(device)

prompt = "girl, smiling, red eyes, blue hair, white shirt"
negative_prompt="low quality, bad quality"
image = pipe(
    prompt=prompt,
    negative_prompt = negative_prompt,
).images[0]

image.show(image)

Model Limitations:

• Limited concept coverage due to the small dataset.

Acknowledgments

• Stan — Key investor. Thank you for believing in us when others called it madness.
• Captainsaturnus
• Love. Death. Transformers.
• TOPAPEC

Datasets

• CaptionEmporium

Donations

Please contact with us if you may provide some GPU's or money on training

DOGE: DEw2DR8C7BnF8GgcrfTzUjSnGkuMeJhg83

BTC: 3JHv9Hb8kEW8zMAccdgCdZGfrHeMhH1rpN

Contacts

recoilme *prefered way

mail at aiartlab.org (slow response)

More examples

Citation

@misc{sdxs,
  title={Simple Diffusion XS},
  author={recoilme with help of AiArtLab Team},
  url={https://huggingface.co/AiArtLab/sdxs},
  year={2025}
}