InvokeAI is an AI image generation creation engine for creative professionals, positioned as a fully featured, professional-grade AI painting tool, with production-grade workflows as its core design philosophy.
In terms of main capabilities, Invoke provides true layer-based canvas editing, allowing users to draw, doodle, sketch, and edit works, with each layer independently operable, making it easy to freely combine complex scenes. Its node-based workflow allows the construction of complex and reproducible generation pipelines through a fully visual graph structure, and can expose custom UI parameters, making it convenient to share and update values without needing to deeply understand the graph structure. In terms of model support, it supports the latest foundation models such as Flux, SDXL, and SD 1.5 out of the box, and manages checkpoints, LoRA, Textual Inversion, and ControlNet through an intuitive model manager. The ControlNet implementation covers multiple control methods such as depth maps, edges, and pose, allowing precise control of composition; it also supports batch generation, prompt wildcards, and high-resolution upscaling, facilitating rapid concept iteration.
In terms of usage, Invoke is fully locally self-hosted, running on the user's own hardware, with data, prompts, and creative content fully belonging to the user, without relying on cloud services. Official versions are provided for Windows, macOS, and Linux. Windows requires Windows 10 or later and an NVIDIA or AMD GPU; macOS requires Apple Silicon (M series) and is not compatible with Intel; Linux requires an NVIDIA or AMD GPU and is compatible with most distributions. Installation can be completed quickly via the Invoke Launcher, and advanced users can also choose Docker or manual Python installation. For users who wish to use hosted GPU services instead of their own hardware, corresponding solutions are also provided.
In terms of applicable scenarios and limitations, Invoke is suitable for professional creators and enthusiasts who need fine layer editing, reproducible node pipelines, multi-model and ControlNet control, batch generation, and local privacy protection. In terms of model training, Invoke Training provides a dedicated UI for Textual Inversion and LoRA training. The project is open-source software, licensed under Apache 2.0, and maintained by an active open-source community. The original official hosted platform has been shut down, the founding team joined Adobe, and the project has been taken over and managed by core maintainers since its founding, Lincoln Stein (lstein) and Vic (Blessedcoolant). The open-source version provides the same functionality as the hosted version, with the additional advantages of full control and privacy.
