- ControlNet - Full support for ControlNet, including custom ControlNets. Set them using
context.model_paths['controlnet'] = '/path/to/controlnet.safetensors'; load_model(context, 'controlnet'). - SDXL - Full support for SDXL. No configuration necessary, just load the SDXL model like any stable diffusion model.
- Multiple LoRAs - Use multiple LoRAs, including SDXL and SD2-compatible LoRAs. Set them using
context.model_paths['lora'] = '/path/to/lora.safetensors'; load_model(context, 'lora'). - Embeddings - Use textual inversion embeddings easily. Set them using
context.model_paths['embeddings'] = '/path/to/embedding.pt'; load_model(context, 'embeddings'). Thanks @JeLuF. - Seamless Tiling - Generate repeating textures that can be useful for games and other art projects. Works best in 512x512 resolution. Thanks @JeLuF.
- Inpainting Models - Full support for inpainting models, including custom inpainting models. No configuration (or yaml files) necessary.
- Faster than v2.5 - Nearly 40% faster than sdkit v1.0, and can be even faster if you enable xFormers.
- Even less VRAM usage - Less than 2 GB for 512x512 images on 'low' VRAM usage setting (SD 1.5). Can generate large images with SDXL.
- WebP images - Supports saving images in the lossless webp format.
- Three new samplers, and latent upscaler - Added
DEIS,DDPMandDPM++ 2m SDEas additional samplers. Thanks @ogmaresca and @rbertus2000. - Major rewrite of the code - We've switched to using diffusers under-the-hood, which allows us to release new features faster.
Contributors
JeLuF, ogmaresca, and rbertus2000