rupeshs/fastsdcpu

Fast stable diffusion on CPU and AI PC

What it solves

FastSD CPU enables high-speed image generation using Stable Diffusion on CPUs, overcoming the typical slowness of CPU-based inference. It allows users to generate images in as little as one step, significantly reducing the time required to create visuals on hardware that lacks a powerful dedicated GPU.

How it works

The project utilizes Latent Consistency Models (LCM) and Adversarial Diffusion Distillation (ADD) to reduce the number of inference steps needed. It leverages Intel's OpenVINO toolkit to further accelerate performance on CPUs, integrated GPUs (iGPUs), and NPUs. To save memory and increase speed, it supports Tiny AutoEncoders (TAESD), which replace the standard decoder to reduce RAM usage and latency.

Who it’s for

It is designed for users who want to run Stable Diffusion locally on Windows, Linux, Mac, Android (via Termux), or Raspberry Pi 4, particularly those without high-end NVIDIA GPUs.

Highlights

  • Multi-Interface Support: Offers a Desktop GUI (Qt), a WebUI for advanced features, and a Command Line Interface (CLI).
  • Extreme Speed: Capable of generating a 512x512 image in 0.82 seconds on a Core i7-12700 using OpenVINO and SDXS-512-0.9.
  • Broad Model Support: Compatible with SD 1.5, SDXL, SD Turbo, SDXL-Lightning, FLUX.1, and FLUX.2-klein.
  • C-Suite Features: Includes support for LoRA, ControlNet, image-to-image, image variations, and image editing (restoration and colorization).
  • Hardware Optimization: Specifically optimized for Intel AI PCs, including support for Lunar Lake NPUs.

Related

  • Dispatch
    Accelerating Stable Diffusion Inference on Intel Sapphire Rapids CPUs

    Hugging Face shows how to cut Stable Diffusion image generation from 32 seconds to about 5 seconds on Intel Sapphire Rapids CPUs using Optimum Intel, OpenVINO, system‑level tweaks, IPEX BF16, and a faster scheduler.

  • Project
    vladmandic/sdnext

    A comprehensive web interface for AI image and video generation based on Stable Diffusion, featuring advanced memory optimizations and broad hardware support.

  • Dispatch
    Optimizing Stable Diffusion for Intel CPUs with NNCF and 🤗 Optimum

    Hugging Face announced a workflow using NNCF and Optimum to quantize Stable Diffusion and apply Token Merging, achieving up to 5.1x inference speedup and a 4x reduction in model footprint on Intel CPUs.

  • Project
    xororz/local-dream

    An Android application that enables local Stable Diffusion image generation with hardware acceleration for Snapdragon NPUs.

  • Dispatch
    Stable Diffusion with Diffusers

    Hugging Face announced how to run the text-to-image Stable Diffusion model using the Diffusers library, providing code, licensing details, and explanations of the model’s latent diffusion architecture.