Small-sd is a text-to-image diffusion model hosted on the Hugging Face Hub. It was created by Segmind through distillation of the Realistic Vision V4.0 model using a subset of the recastai/LAION-art-EN-improved-captions dataset. The result is a smaller and faster model that maintains image generation quality from text prompts.
The model is provided as a StableDiffusionPipeline compatible with the Diffusers library. It supports inference with bfloat16 precision and can run on CUDA devices or Apple MPS hardware. Example code demonstrates loading the pipeline from the repository identifier segmind/small-sd and generating an image from a descriptive prompt such as "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k".
It is distributed under the creativeml-openrail-m license. The associated paper is arxiv:2305.15798. Local applications that can use the model include Draw Things and DiffusionBee. Notebooks for Google Colab and Kaggle are also referenced for getting started.
The model card lists integration with Diffusers and provides instructions for installation via pip along with the required transformers and accelerate packages. It belongs to the class of distilled text-to-image models intended for more accessible local inference on consumer hardware.
In the Image generation space, Small Sd takes a focused approach. It focuses on generating images from text prompts with a smaller, faster Stable Diffusion variant. It is built as an open-source project for AI developers. Small Sd is open source under the Open Source license. Small Sd is available on the web, the command line, and API.
Behind Small Sd is Segmind, and the product first shipped in 2023. Development happens publicly on GitHub with 618 stars. Key capabilities include text-to-image, distilled model, and diffusers support.
Latest indexed changes and source events
segmind/small-sd verified by the PulseGate indexer
Other apps tracked under the same category.
