InvokeAI (Invoke) is a self-hosted engine for AI image generation that runs a local web server with a React interface. It is aimed at artists and production teams and centres on two working modes: a layer-based Unified Canvas for drawing, painting, inpainting, and outpainting alongside generated imagery, and a node graph for building reproducible generation pipelines that can expose their own parameters and be shared.
Generated images are stored in boards and a gallery, carry metadata that recalls the prompts and settings used, and can be dragged onto any image-accepting element in the interface. A model manager handles checkpoints, LoRAs, textual inversions, and ControlNets in ckpt, diffusers, and some GGUF formats.
Features
- Unified Canvas: layer-based editing with brush tools, in/out-painting, and every core generation capability on one surface
- Workflows and nodes: visual graphs for custom pipelines with exposed UI parameters for sharing and reuse
- Boards and gallery: organised storage with drag-and-drop and per-image metadata for recalling prompts and settings
- Local models: SD 1.5, SD 2.0, SDXL, SD 3.5 Medium and Large, CogView 4, Flux.1 Dev, Schnell, Kontext, and Krea, Flux Redux and Fill, Flux.2 Dev and Klein 4B/9B, Z-Image Turbo and Base, Krea 2 Turbo and Raw, Anima, Qwen Image and Qwen Image Edit, Ideogram 4, and ERNIE-Image
- API-only models: Nano Banana, GPT Image, and Wan
- Model manager: checkpoints, LoRAs, textual inversions, and ControlNets, with ckpt, diffusers, and some GGUF formats
- Segmentation: object selection with SAM and SAM2 models
- Upscaling: built-in upscaling tools
- Local and private: runs entirely on your own hardware with no external service required
