ComfyUI
GPL-3.0 node graph for image, video, audio and 3D, with Stable Diffusion, SDXL, Flux, Wan and LTX, LoRAs and ControlNets, running fully offline.
Know firstA node graph takes learning, and commits between stable tags can break custom nodes.
Image and video generation4 projects, checked September 17, 2026
Open-source alternatives to Midjourney
The open-source alternatives to Midjourney are ComfyUI, InvokeAI, SD.Next and SwarmUI. Each runs open image models such as Stable Diffusion, SDXL and Flux on your own hardware, and three of them also generate video. ComfyUI gives a node graph with the widest model support, InvokeAI a canvas for inpainting and outpainting, SD.Next a form-based interface that runs on modest or unusual GPUs, and SwarmUI a simple tab with ComfyUI's full graph behind it. The software is free; each model has its own licence.
- Control, and the newest models
- ComfyUI
- A canvas for inpainting and editing
- InvokeAI
- Modest or unusual hardware
- SD.Next
4 of 4 carry an OSI licence. 4 of 4 run on your own machines. Figures read from GitHub on September 17, 2026.
Pick by need
One project per need, with the reason from its own README and the one thing to know before you choose it.
InvokeAI
Apache-2.0 with a unified canvas for generation, inpainting, outpainting and brushes, node workflows, and 28 model families.
Know firstHardware requirements are not stated, and three listed models run only through paid APIs.
SD.Next
Apache-2.0 web interface with LoRA, ControlNet and quantisation its README says cuts video memory up to four times, on AMD, Intel, Apple or DirectX GPUs and CPUs.
Know firstThe README lists no model families and the project has no GitHub releases, only dated tags.
SwarmUI
MIT interface with a beginner Generate tab and the raw ComfyUI graph behind it, plus multi-GPU generation for one user.
Know firstIts README calls it Almost-Release status, and the last tagged release is a beta.
What to take away
- The interfaces are open source; each model you download has its own licence, and some forbid commercial use.
- ComfyUI has the widest model support and runs fully offline.
- SD.Next is the one built for hardware that other tools struggle on.
- Midjourney still wins on needing no hardware, no downloads and no setup.
The projects, side by side
The facts that decide most choices. Each project's page carries the full panel.
| Project | Best for | Openness | Licence | Runs as | Own machines | Stars |
|---|---|---|---|---|---|---|
| ComfyUI | Control, and the newest models | OSI | GPL-3.0 | Desktop app, Portable build, pip (comfy-cli), From source | Yes | 133,702 |
| InvokeAI | A canvas for inpainting and editing | OSI | Apache-2.0 | Launcher (desktop), Local web server | Yes | 28,237 |
| SD.Next | Modest or unusual hardware | OSI | Apache-2.0 | Web UI (built-in installer), Docker | Yes | 7,342 |
| SwarmUI | A simple tab now, a graph later | OSI | MIT | Web UI (installer scripts), Docker, Google Colab | Yes | 4,571 |
Licence and star figures read from GitHub on September 17, 2026. Stars are shown as a dated fact and were not used to rank this page.
Each project, in turn
The GPL-3.0 node-graph engine for image, video, audio and 3D generation that supports the newest open models, from Stable Diffusion and Flux to Wan and LTX video, on NVIDIA, AMD, Intel and Apple hardware, offline.
GPL-3.0 node graph for image, video, audio and 3D, with Stable Diffusion, SDXL, Flux, Wan and LTX, LoRAs and ControlNets, running fully offline.
Know firstA node graph takes learning, and commits between stable tags can break custom nodes.
- Licence
- GPL-3.0
- Runs as
- Desktop app, Portable build, pip (comfy-cli), From source
- Platforms
- Windows, macOS (Apple Silicon), Linux
- Hosted option
- Sold by the maintainers
- Last release
- v0.36.0, Sep 15, 2026
An Apache-2.0 creative image tool with a unified canvas for inpainting and outpainting, node workflows, and support for 28 model families from Stable Diffusion to Flux and Qwen Image, run as a local web app.
Apache-2.0 with a unified canvas for generation, inpainting, outpainting and brushes, node workflows, and 28 model families.
Know firstHardware requirements are not stated, and three listed models run only through paid APIs.
- Licence
- Apache-2.0
- Runs as
- Launcher (desktop), Local web server
- Platforms
- Not stated in the README
- Hosted option
- None from the maintainers
- Last release
- v6.14.1, Sep 6, 2026
An Apache-2.0 all-in-one web interface for AI image and video generation, descended from AUTOMATIC1111's WebUI, with LoRA, ControlNet, quantisation that cuts video memory up to four times, and the widest hardware list here.
Apache-2.0 web interface with LoRA, ControlNet and quantisation its README says cuts video memory up to four times, on AMD, Intel, Apple or DirectX GPUs and CPUs.
Know firstThe README lists no model families and the project has no GitHub releases, only dated tags.
- Licence
- Apache-2.0
- Runs as
- Web UI (built-in installer), Docker
- Platforms
- Windows, Linux, macOS (Apple M1 and M2), WSL
- Hosted option
- None from the maintainers
- Last release
- No tagged release
An MIT image and video generation web interface, formerly StableSwarmUI, with a simple Generate tab for beginners and a raw ComfyUI workflow tab for experts, that can spread work across several GPUs.
MIT interface with a beginner Generate tab and the raw ComfyUI graph behind it, plus multi-GPU generation for one user.
Know firstIts README calls it Almost-Release status, and the last tagged release is a beta.
- Licence
- MIT
- Runs as
- Web UI (installer scripts), Docker, Google Colab
- Platforms
- Windows 10 and 11, Linux, macOS (M-series), Docker
- Hosted option
- None from the maintainers
- Last release
- 0.9.8-Beta, Feb 6, 2026
How to choose
The questions that settle it, in the order they usually come up.
How much control do you want?
ComfyUI's graph for every step, SwarmUI's simple tab to start, InvokeAI's canvas for editing, SD.Next's form for a familiar layout.
What hardware do you have?
SD.Next covers AMD, Intel, Apple and DirectX GPUs and CPUs with memory-saving quantisation. ComfyUI covers NVIDIA, AMD, Intel, Apple and Ascend.
Do you need video?
ComfyUI, SwarmUI and SD.Next generate video; InvokeAI lists one video model as API only.
Where Midjourney is still the better choice
Midjourney is not described here in its own words: its site refused every automated request on September 17, 2026, and this directory does not write a description it has not read. See midjourney.com for the vendor's own account. The cases below are where it still wins.
- You want images from a sentence with no hardware, no model downloads and no setup.
- You want one consistent house style out of the box, which open models reach only with work.
- You want a community gallery and shared prompts as part of the product.
- You do not want to track which model licence allows commercial use; that question is yours with open models.
How this page was checked
Numbers from GitHub, dated
Stars, licence, last commit and last release are read from the GitHub API by a script every week and carry the date they were read. Nobody types them.
Claims from the README
What each project does is written from its own README and LICENSE, read in full on the date shown, and phrased as the project's claim. The sources are listed at the end.
What is never said
That a project is faster, better or cheaper than Midjourney. A price for anyone. A user count. A roadmap. If a fact is not on this page, it was not in the source.
- OSI Open source (OSI licence)
- The code, and the weights where the project is a model, carry a licence on the OSI approved list.
- Source available Source available
- The code is public, and the licence limits what you may do with it. Read the LICENSE file before commercial use.
- Open weights Open weights
- The model weights can be downloaded and run. The training data or code is closed, or the weights carry a restriction the code does not.
Questions people ask
What is the best open-source alternative to Midjourney?
ComfyUI (GPL-3.0) for the widest support of new open models and full control through a node graph, including video. InvokeAI (Apache-2.0) for a canvas with inpainting and outpainting. SD.Next (Apache-2.0) for AMD, Intel or low-memory hardware. SwarmUI (MIT) for a simple interface with ComfyUI behind it.
Can I sell images made with these?
The software licences allow it, but the model does not always. Each open model carries its own licence and some are non-commercial, as SwarmUI's README notes. Check the model card of whichever model you generate with, not just the interface you run it in.
What hardware do I need?
It depends on the model. SD.Next lists NVIDIA, AMD, Intel, Apple and DirectX GPUs and CPU-only use, with quantisation that its README says cuts video memory up to four times. ComfyUI covers NVIDIA, AMD, Intel, Apple Silicon and Ascend. None of the READMEs states a minimum amount of video memory.
Can these make video as well as images?
Yes, three of them. ComfyUI lists Wan, LTX-Video, HunyuanVideo, CogVideoX and Mochi. SwarmUI lists MiniMax H3, Wan and LTX-2. SD.Next covers text to video and image to video. InvokeAI lists one video model, Wan, as API only.
Why is there no description of Midjourney here in its own words?
Because its site refused every automated request when this page was written, and this directory does not write a description it has not read. Every other page here quotes the vendor's own words from a page that could be fetched.
Sources
- ComfyUI README and GPL-3.0 licence, read September 17, 2026
- InvokeAI README, read September 17, 2026
- SD.Next README, read September 17, 2026
- SwarmUI README, read September 17, 2026
Running one of these inside your own environment, with your own data and your own security rules, is the kind of work Reveneau does. Read how a forward deployed engagement works.