Image generation

June 25, 2026 · View on GitHub

← All features

  • Text → image and image → image via stable-diffusion.cpp (Metal-accelerated).
  • Works with SDXL finetunes as GGUF (RealVis, Juggernaut, DreamShaper, …) and other checkpoints — pick the active image model in Models.
  • Curated styles with on-device preview thumbnails; LoRA support.
  • Tune resolution, steps, seed; live progress with previews. Outputs saved to the gallery (scoped to the chat/project) and openable from chat.