InvokeAI
invoke-ai/InvokeAIReviewed by AI
Checked by RepoPicks Radar on
What is InvokeAI?
Invoke is a free program for creating and editing images with AI models that run on your own computer. You install it with its launcher, download the models you want inside the app, and work in a browser window on your machine: a prompt box, a paint canvas for filling in or extending parts of a picture, galleries and boards, and node workflows for repeatable jobs. It supports many model families, from Stable Diffusion to FLUX and Qwen Image, and a few online-only models. It is aimed at artists and studios and needs a capable graphics card.
InvokeAI is free and open source (Apache-2.0) and can replace Midjourney, Adobe Firefly, and Adobe Photoshop.
How easy is it to install InvokeAI?
B · installer plus one simple step · Windows
- Windows B
- macOS B
- Linux B
On Windows, Mac and Linux the README points at one download, "Download the Launcher", which sets the program up for you; after that the extra step is picking and downloading models inside the app before you can generate anything. There is no terminal work on the normal route, but you do need a suitable graphics card, and the pyproject file shows the Mac build is held back to older graphics libraries because newer ones run out of memory.
InvokeAI on GitHub
- Stars
- 29,054on Nov 10, 2026
- Downloads
- 355,026all releases
- Growth
- —no gain in the last reading
Stars over time
History starts on Nov 10, 2026.
Facts
- Runs on
- Windows, macOS, Linux
- You need to install
- NVIDIA graphics card
- To build from source
- Python, NVIDIA graphics card, FFmpeg
- Disk space
- 23 MB–69 MB · from the installer
- Accounts or keys
- Online image and video model services for the API-only models (Nano Banana, GPT Image, Wan)
- Cost
- Free · The program is free under the Apache-2.0 licence. The company behind it asks for sponsorship and says Invoke "serves as the foundation for multiple commercial products", and a few models in the list work only through a paid online service ("Nano Banana (API Only)", "GPT Image (API Only)", "Wan (API Only)").
- License
- Apache-2.0 (Permissive)
- Can replace
- Midjourney, Firefly, Photoshop
- Last change on GitHub
- Nov 10, 2026
InvokeAI risk signals (5)
- Attentionthe models have their own licences
The program is Apache-2.0, but the repository also carries LICENSE-SD1+SD2.txt, LICENSE-SDXL.txt, LICENSE-HiDiffusion.txt, LICENSE-PiD.txt and Stable_Diffusion_v1_Model_Card.md, and the model list includes families such as "Flux.1 Dev" and "Flux.2 Dev". The models you download come under those separate terms, which can limit commercial use, and the app's licence does not cover them.
AI review
- Infothe installer lives in another repository
This repository's three newest releases (v7.0.0-beta.1, v6.14.4, v7.0.0-alpha.4) have "no platform binaries"; the README sends you to a different project for the install: "To get started with Invoke, [Download the Launcher](https://github.com/invoke-ai/launcher/releases/latest)". That code is not in this repository.
AI review
- Infothe newest release is a test version
The latest release is v7.0.0-beta.1 (2026-11-09) and the one before it is v7.0.0-alpha.4; the newest finished version is v6.14.4 (2026-11-03). The project has 235 releases and 318 open issues.
AI review
- Infomodels are downloaded and take a lot of disk
The app ships no models: it has a "Model Manager & Support" and the model list names more than 25 families to fetch, from SD 1.5 to Flux.2 and Qwen Image. The size the checks report (23 to 69 MB) is only the program; each model you download is typically several gigabytes.
AI review
- Infoit opens as a web page served by your own computer
README: "Invoke runs a locally hosted web server & React UI". You use it in a browser against a server running on your machine, so anything that can reach that port can reach the app.
AI review
Evidence is shown in its original language.
What Radar could not check about InvokeAI
- Which graphics cards are enough for each model family, and how much video memory the newer FLUX and Qwen models need.
- Real disk use once you download the models you want, on top of the program itself.
- How well it performs on a Mac, where the project pins older graphics libraries, and on AMD cards on Linux.
- Whether the separate launcher download installs what this repository contains.
- What the API-only models cost per image or video, since that is billed by the outside service.
- Whether the local web server is reachable from other machines on your network in the default setup.