SDXL-Turbo
A few-step distillation text-to-image model developed by Stability AI, serving as the fast-inference variant of SDXL. Through distillation techniques, it substantially reduces the number of required sampling steps, enabling efficient image generation with significantly lower computational cost than the original SDXL model.
Timeline (last 90 days)
Representative few-step distilled models include SDXL-Turbo, LCM (Latent Consistency Model), and Flash Diffusion.
InstructPix2Pix, SDXL-Turbo, and others explored a unified generation and editing framework, but there was often a trade-off between generation quality and editing precision.
SD-Turbo/SDXL-Turbo introduced by Stability AI achieves single-step generation through distillation.
Common inference acceleration techniques include consistency distillation, flow matching, and diffusion model step compression methods such as LCM and SDXL-Turbo.
All Facts (4)
Representative few-step distilled models include SDXL-Turbo, LCM (Latent Consistency Model), and Flash Diffusion.
50%UnverifiedInstructPix2Pix, SDXL-Turbo, and others explored a unified generation and editing framework, but there was often a trade-off between generation quality and editing precision.
50%UnverifiedSD-Turbo/SDXL-Turbo introduced by Stability AI achieves single-step generation through distillation.
50%UnverifiedCommon inference acceleration techniques include consistency distillation, flow matching, and diffusion model step compression methods such as LCM and SDXL-Turbo.
50%