InvokeAI Open source

What it is

InvokeAI is an open-source image generation studio (Apache-2.0, repo invoke-ai/InvokeAI) with a layered Unified Canvas, a gallery and a node-based workflow editor. Its commercial company shut down the hosted Invoke platform on 31 October 2025 when key staff joined Adobe; the open-source Community Edition continues under longtime maintainers and still ships regular releases (v6.14.x in 2026). You install it locally and generate images, edits and, in recent releases, short videos.

Key features

• Unified Canvas with layers, regional prompts, inpainting and outpainting
• Node-based workflow editor alongside the simple prompt UI
• Model manager for checkpoints, LoRAs and ControlNets
• Recent releases add FLUX.2, Qwen Image, Z-Image and FP8 support to save VRAM
• Optional hosted-model integrations such as GPT Image, Gemini and Seedream
• Launcher installer for Windows, macOS and Linux

Models inside

Locally: Stable Diffusion 1.5 and SDXL, FLUX (including FLUX.2 variants), Qwen Image and Qwen Image Edit, Z-Image and other open models listed in the release notes, with Wan 2.2 video noted in recent releases. External API integrations in the release notes include OpenAI GPT Image, Google Gemini image models and ByteDance Seedream, billed by those providers.

Pricing and rights

InvokeAI is free under Apache-2.0 and runs on your own GPU with no watermark and no usage limits; commercial use depends on each model's licence. There is no longer a paid hosted plan from Invoke, so there is no subscription and no vendor support contract.

Who it is for

Illustrators, designers and hobbyists who want a polished, canvas-first local image editor rather than a raw node graph, and who have a capable GPU.

Verdict

InvokeAI keeps one of the best local canvas experiences and still gains new models, but it now relies on volunteer maintainers after the company's exit. Expect slower support and fewer integrations than ComfyUI.