What it is
InvokeAI is a local application for image generation with Stable Diffusion, FLUX, Qwen Image, Z-Image and other open models, plus API-only models such as GPT Image and Nano Banana. It offers a layered canvas for inpainting and editing, a gallery and a node-based workflow editor, served as a local web UI.
In October 2025 the founding team joined Adobe and the hosted Invoke service shut down; the open-source project continues under long-time maintainers Lincoln Stein and blessedcoolant, funded by GitHub sponsors. The stable line is v6.14.2 from 27 September 2026.
A v7 alpha began in October 2026 with a rewritten canvas, projects, semantic gallery search and video generation with Wan 2.2, LTX-2.5 and MiniMax H3. It needs Python 3.12 and migrates the database one way, so a v6 install cannot open it afterwards.
Our review
InvokeAI survived the loss of its company better than most expected, but anyone installing it in late 2026 is choosing between a stable v6 and an alpha v7 that changes almost everything. It has about 28,800 GitHub stars and its layered canvas, with masks, inpainting and regional guidance, is still closer to a design tool than ComfyUI's node graph.
After the founders left for Adobe and the hosted service closed in October 2025, maintainers kept shipping: v6.14.2 came out on 27 September 2026, and the model list now runs from SD 1.5 to FLUX.2 Klein, Qwen Image and Krea 2. The hardware guide is honest about cost: SDXL wants an 8 GB NVIDIA card and 16 GB of RAM, FLUX.1 10 GB and 32 GB, and FLUX.2 Klein 9B or Krea 2 a 24 GB RTX 40-series card unless you drop to FP8 or GGUF. AMD works on Linux only.
v7, in alpha since 4 October 2026, is the biggest release in the project's history at about 320 pull requests. It rewrites the canvas with layer groups, 16 blend modes and PSD export, adds video with sound, and makes the new interface the default. The release notes warn that it will break, that it upgrades the database irreversibly, and that it should run in a separate root directory. Video brings it nearer ComfyUI, which has supported those models for longer.
The question for a studio is less features than staffing. A volunteer project with no commercial backer and 383 open issues and pull requests is about to ask every user to migrate, and new model support depends on contributors. ComfyUI's larger ecosystem absorbs that risk better; InvokeAI's advantage is that a retoucher can use it without learning a graph.
Where it earns its place
A freelance illustrator with a 12 GB RTX 3060
Inpaint and extend client artwork locally without uploading it
Install with the Launcher and use SDXL or FLUX.2 Klein 4B on the canvas, inside the published VRAM limits. No subscription, and images stay on the machine; heavier models need FP8 or GGUF builds.
A five-person design studio sharing one 24 GB GPU workstation
Give staff a shared image tool without paying for per-seat cloud credits
Run one Invoke server and enable its documented multi-user mode, which has separate user and administrator guides. Hardware is the only cost; someone must own updates and decide when to move to v7.
A creative director comparing local tools after Invoke's hosted service closed
Replace a cancelled cloud subscription with something the team can run
Self-hosted InvokeAI keeps the familiar canvas but has no vendor support contract. ComfyUI offers more models and nodes; Invoke trades that for a gentler learning curve for non-technical staff.
Before you commit
- Setup
- Download the Invoke Launcher for Windows, macOS or Linux, which installs Python 3.11 to 3.12 and the app; Docker and manual pip installs also exist. v7 alpha needs Python 3.12 and a fresh root directory.
- Integrations
- Node-based workflows with community node packs, a REST API for saved workflows, a ComfyUI migration guide, and external API models (OpenAI, Google Gemini, Alibaba DashScope, BytePlus Seedream).
- Data and privacy
- Runs locally and stores images and metadata in a local SQLite database; only the optional external-model nodes send prompts to third-party APIs. No GitHub security advisories published.
- Cost at scale
- Free software; the cost is hardware. Per the docs, SDXL needs an 8 GB NVIDIA GPU and 16 GB RAM, FLUX.1 10 GB VRAM and 32 GB RAM, FLUX.2 Klein 9B and Krea 2 24 GB VRAM and 32 GB RAM; Apple Silicon wants 16 GB or more.
- Limits
- 383 open issues and pull requests; AMD GPUs on Linux only; v7 alpha breaks v6 databases; new open models tend to arrive later than in ComfyUI.
What's good
- Layered canvas suited to artists and retouchers
- Launcher installs Python and dependencies for you
- Published VRAM and RAM guide per model family
- Releases continued after the company left
What's not
- No company behind it; volunteer maintainers
- v7 alpha migrates the database one way
- Large models want a 24 GB GPU
- Smaller extension ecosystem than ComfyUI
Best for
Artists who want canvas-style image generation and editing on their own machine.
Alternatives
InvokeAI: common questions
- Is InvokeAI worth it?
- An artist-friendly local image app that has outlived its company, now mid-way through a risky v7 rewrite. We rate it 3.9 out of 5.
- How much does InvokeAI cost?
- Free and open source under the Apache 2.0 licence. No paid tier since the hosted service closed in October 2025.
- What are the downsides of InvokeAI?
- No company behind it; volunteer maintainers; v7 alpha migrates the database one way; large models want a 24 GB GPU; smaller extension ecosystem than ComfyUI.
- Who is InvokeAI best for?
- Artists who want canvas-style image generation and editing on their own machine.
- How much does InvokeAI cost for a team or in production?
- Free software; the cost is hardware. Per the docs, SDXL needs an 8 GB NVIDIA GPU and 16 GB RAM, FLUX.1 10 GB VRAM and 32 GB RAM, FLUX.2 Klein 9B and Krea 2 24 GB VRAM and 32 GB RAM; Apple Silicon wants 16 GB or more.
- Is InvokeAI safe for company data?
- Runs locally and stores images and metadata in a local SQLite database; only the optional external-model nodes send prompts to third-party APIs. No GitHub security advisories published.
- What are the best InvokeAI alternatives?
- The closest alternatives we have reviewed are ComfyUI and Hugging Face Diffusers.
Answers checked Oct 23, 2026.