selfhost.directory

The index of self-hostable projects — ranked, categorized and served with live updates.
Own your data and stop giving it to big tech.

InvokeAI logo

Self-host InvokeAI

v6.14.4

Creative engine for Stable Diffusion and Flux generation

ReleaseOpen SourceApache-2.0
docker run --runtime=nvidia --gpus=all --publish 9090:9090 ghcr.io/invoke-ai/invokeai
Updatedlast week
LanguageTypescript
Setup difficultyEasy
Stars
29 097
RAM
8GB
Port
9090 (InvokeAI web server and UI (localhost:9090 by default))
Docker
ghcr.io/invoke-ai/invokeai

What is InvokeAI?

Invoke is an Apache-2.0 licensed creative engine for AI image generation, aimed at professionals, artists and enthusiasts who want to work locally. It ships a web server and React UI whose centrepiece is a layered canvas supporting in- and out-painting, brush, shape, text and lasso tools, alongside a node-based workflow editor for building and sharing reproducible generation pipelines. Model coverage is broad and current - Stable Diffusion 1.5/2.0, SDXL, the SD3.5 sizes, the Flux.1 and Flux.2 families, CogView 4, Qwen Image and others - in checkpoint, diffusers and some gguf formats, with a model manager, embedding manager, boards and gallery, upscaling, batch runs, prompt wildcards and extensive ControlNet guidance. Opt-in multi-user mode adds accounts, an admin panel and round-robin queue scheduling for shared machines, and the backend can bind to 0.0.0.0 for LAN use behind a reverse proxy. Most installs use the Invoke Launcher on Windows, macOS or Linux; a ghcr container publishing port 9090 and a manual Python 3.11-3.12 install are also documented. Big models need a GPU with ample VRAM, AMD support is Linux-only, and ARM64 Linux runs CPU-only. The hosted Invoke.ai service was retired after the founding team joined Adobe, so the community-maintained self-hosted app is now the only way to use it.

Should you self-host InvokeAI?

InvokeAI scores 81/100 on selfhost.directory (how scoring works) and ranks #2 of 11 in Image Generation. It has 29 097 GitHub stars, +530 in the last 30 days.

It shipped 15 stable releases in the past year. The latest, v6.14.4, came out on 3 November 2026. New versions land roughly every 2 weeks.

Setup is easy: most people get it running in one sitting. An official Docker image is available. It is Apache-2.0-licensed and written mostly in Typescript.

Its closest alternative on selfhost.directory is stable-diffusion-webui (89/100, 165 318 stars). It scores higher overall, so compare the two before you pick. See InvokeAI vs stable-diffusion-webui side by side.

Features

  • API-backed external modelsrequires external provider API keys
  • Batch Generation
  • Broad local model support
  • ControlNet
  • Gallery and board management
  • LAN / reverse-proxy serving
  • Local-only generation
  • Model manager
  • Multi-User Mode
  • Node-based workflows
  • Segmentation models (SAM / SAM2)
  • Upscaling

InvokeAI README

project hero

Invoke - Professional Creative AI Tools for Visual Media

discord badge latest release badge github stars badge github forks badge latest commit to main badge github open issues badge github open prs badge translation status badge

Sponsor Invoke

How to install InvokeAI

Install InvokeAI with Docker

Environment variables
INVOKEAI_ROOT, HUGGING_FACE_HUB_TOKEN, GPU_DRIVER, INVOKEAI_PORT, HOST_INVOKEAI_ROOT, CONTAINER_INVOKEAI_ROOT, RENDER_GROUP_ID, CONTAINER_UID
docker run --runtime=nvidia --gpus=all --publish 9090:9090 ghcr.io/invoke-ai/invokeai

Once the container starts up, open http://localhost:9090 in your browser, install some models, and start generating.

Build-it-yourself: all the docker materials are located inside the docker directory in the Git repo:

cd docker
cp .env.sample .env
docker compose up

Configure Docker to access the GPU (NVIDIA container toolkit or ROCm for AMD). Docker cannot access the GPU on macOS.

How to configure InvokeAI

  • INVOKEAI_ROOT~/invokeaiLocation of the InvokeAI root directory (models, config, outputs); created if missing. Used as the docker host data path.
  • INVOKEAI_HOST127.0.0.1IP address to bind to; use 0.0.0.0 to serve to the local network. YAML key: host
  • INVOKEAI_PORT9090Port on which the InvokeAI web interface is available.
  • INVOKEAI_MODELS_DIRmodelsPath to the models directory (YAML key: models_dir).
  • INVOKEAI_OUTPUTS_DIRoutputsDirectory for generated outputs (YAML key: outputs_dir).

InvokeAI releases & version history

@ Every self-hosted release, weekly

One Monday email with the week's stable releases across the directory.

FAQ

What's the licensing impact of multi-user mode?

None - the multi-user feature does not change the project's existing license terms.

How do I handle user data when they leave?

Deactivate the account first, hand over shared boards, then delete it (or leave it deactivated) after a transition period.

Can users have different model access?

Not currently - every user can see and use all installed models, and per-user model access is only a possible future enhancement.

How do I audit user actions?

Full audit logging is a future release item; for now admins watch the generation queue, review database changes and check application logs.

Can I integrate with existing authentication systems?

No - OAuth2/OpenID Connect is only planned, and the app currently relies on its own built-in user accounts.

How many users can InvokeAI support?

The backend handles dozens of concurrent users, but the generation queue runs one job at a time, scheduled round-robin per user by default (session_queue_mode: FIFO switches to first-come, first-served).

Models InvokeAI runs

Popular open-weight models with InvokeAI support, with VRAM needs and licence at a glance.

black-forest-labs logo
FLUX.1-Kontext-devblack-forest-labs
11.9BNon-commercial≈15 GB↓ 214.4K
xinsir logo1.3BCommercial OK≈3.3 GB↓ 213.4K
Shakker-Labs logo2.1BNon-commercial≈5.3 GB↓ 18.6K
stable-diffusion-v1-5 logo
stable-diffusion-inpaintingstable-diffusion-v1-5
?BConditions? GB↓ 252.8K
stabilityai logo2.3BConditions≈5.5 GB↓ 111.8K
xinsir logo1.3BCommercial OK≈3.3 GB↓ 296.8K
xinsir logo1.3BCommercial OK≈3.3 GB↓ 62.7K
InstantX logo?BNon-commercial? GB↓ 7.8K

Browse all 43 models →

More in Image Generation

Rate this project

Click to rate — no account needed.

0 comments

Posting as guest
Loading comments…