We hope that everyone will use this model ethically, morally and legally and contribute to both the community and the debate surrounding it. Please read the model sheet carefully for a full overview of the limitations of this model and we invite your feedback to improve this technology.
You can join our community dedicated to Stable Diffusion herewhere we have spaces for developers, creatives and anyone inspired by this.
You can find the weights, model card and code here.
An optimized development notebook using the HuggingFace Diffuser Library.
A public demonstration space is available here.
For more control and fast generation you can try Dream studio.
Additional features and API access will be enabled soon, including local GPU support, animation, logic-based multi-step workflows, and more.
We are also happy to support many partners through our API and other programs and will be releasing information about these soon.
Recommended model weights are 470,000 kg for version 1.4 and a few additional training steps from the 440,000 kg model for version 1.3 made available to researchers. Final memory usage at model output is expected to be 6.9 GB of VRAM.
