Alternatives · Last reviewed October 20, 2026

Invoke (InvokeAI) alternatives, hosted and self-hosted

Invoke (InvokeAI, at invoke.ai) is now free, open-source software you install yourself: a creative engine with a layered canvas, node workflows and gallery boards. Its hosted paid platform has shut down, because the founding team joined Adobe, and two community maintainers now steward the project. People look elsewhere mainly because there is no longer a hosted Invoke, and running it locally needs a recent Nvidia card or an Apple Silicon Mac. ComfyUI is the closest fit for both local and cloud use, Krea for a hosted canvas, and AUTOMATIC1111 or Fooocus for older GPUs. ImageQuarry fits people who want to change an image by describing it without installing anything.

We make ImageQuarry, one of the alternatives below. Invoke (InvokeAI) and each alternative were compared on hosted or self-hosted, hardware needed, editing by describing a change, model choice and price, using each vendor's public pages, checked October 20, 2026.

Start creating See how it works

Last updated October 20, 2026

Invoke (InvokeAI) alternatives at a glance

Tool	Best for	Price from	How it differs
ComfyUI	Invoke users who want node workflows both on their own machine and in a paid cloud	Free locally; Comfy Cloud Standard USD 20 a month, billed monthly (as of 2026-10)	Open-source node graph locally, plus the paid Comfy Cloud
Krea	Former hosted Invoke users who want a browser canvas and node editor with no install	100 free units a day; Basic $9 per month (as of 2026-10)	Hosted real-time canvas, node editor and typed edits
ImageQuarry	Changing an image by describing it, with no install and no layers to manage	From $9 a month (Starter); Pro $29 a month (as of 2026-10)	Describe changes in words; each one becomes a version you can branch
AUTOMATIC1111	Tinkerers with older or smaller GPUs who want a free local web UI	Free and open source (as of 2026-10)	Free self-hosted web UI that supports 4GB video cards
Fooocus	Beginners who want the simplest free local SDXL generator	Free (non-commercial open source) (as of 2026-10)	Prompt-first offline generator on 4GB Nvidia cards

What is Invoke (InvokeAI)?

Invoke describes itself as "a free and open-source creative engine" under the Apache 2.0 license. You install it with a desktop launcher for Windows, macOS on Apple Silicon or Linux, or through Docker, and run Flux, SDXL and SD 1.5 models with LoRAs and ControlNets. The canvas works in layers: raster layers, control layers, regional references and inpaint masks. Images live in Boards, and Canvas Projects save a whole setup into a shareable.invk file. Version 7 is in alpha releases through October 2026.

Invoke at invoke.ai is the InvokeAI image software, not the invoke.com domain, which is parked and for sale.

What Invoke (InvokeAI) does well

Free and open source

Invoke costs nothing and is released under Apache 2.0, with releases still shipping: v7.0.0 alpha 3 came out on October 19, 2026.

A layered canvas for precise control

Dropping an image onto the canvas offers a raster layer, control layer, regional reference or inpaint mask, so you decide exactly which part of the image each input affects.

Local models plus outside ones

The Model Manager runs Flux, SDXL and SD 1.5 checkpoints with LoRAs and Textual Inversions, and external model support sends jobs to Gemini, OpenAI, Seedream and Qwen models billed to your own provider account.

Your settings travel with each file

Every image stores its prompt, seed and models inside the file and can reload them, and a whole board downloads as a.zip.

Shared use on one machine

Multi-User mode, added in version 6.12, lets several people share one Invoke server with separate logins, boards and workflows.

Where Invoke (InvokeAI) falls short

The hosted service is gone

invoke.ai says the hosted platform has been shut down as the founding team joined Adobe, so nothing is sold and you run Invoke yourself or on a third party GPU host.

Real hardware needed

Guideline minimums are an Nvidia 20xx or newer card with 8GB VRAM and 16GB RAM for SDXL, or 10GB VRAM and 32GB RAM for FLUX.1. AMD GPUs work on Linux only, and on Mac only Apple Silicon is supported.

Describing a change depends on outside models

The canvas docs describe layers and masks, with no instruction-style editing apart from external edit models. Inpainting is not supported by any external provider.

Multi-user is for local networks

The Multi-User mode docs warn against exposing the server to the internet, so sharing beyond one network takes extra setup of your own.

Why people look for a Invoke (InvokeAI) alternative

You used the hosted Invoke platform, which has shut down, and you want something you can use in a browser without installing it.
Your computer does not meet the guideline minimums, such as an Nvidia card with 8GB VRAM for SDXL, or you have an Intel Mac.
You want to change an image by describing what should be different, and in Invoke that works only through external models billed to your own provider account, which cannot inpaint.
You want a company that sells and supports the product, and Invoke is now a community project with nothing for sale.

The 5 Invoke (InvokeAI) alternatives

Ordered by how well each covers what Invoke users have lost or lack: a hosted option first, then changing an image by describing it, then running locally on modest hardware.

1. ComfyUI

Best for: Invoke users who want node workflows both on their own machine and in a paid cloud

How it differs from Invoke (InvokeAI): ComfyUI is open source like Invoke but also sells Comfy Cloud, which fills the gap left by Invoke's closed hosted platform. Its Comfy Agent builds and runs workflows from a description, and it runs instruction edit models such as Qwen Image Edit and Flux Kontext. Locally it claims to run large models on as little as 4GB VRAM.

Price: Free locally; Comfy Cloud Standard USD 20 a month, billed monthly (as of 2026-10)

Limitation: Node graphs take time to learn, and Comfy Cloud caps each workflow run at 30 minutes (1 hour on Pro).

Visit ComfyUI

2. Krea

Best for: Former hosted Invoke users who want a browser canvas and node editor with no install

How it differs from Invoke (InvokeAI): Krea is fully hosted, so there is nothing to install, and it has both a live canvas that renders as you draw and a Node Editor for chaining models and steps, two ideas Invoke users will recognize. You can also edit by typing, and it lists Flux, Nano Banana, GPT Image 2.5 and its own Krea models.

Price: 100 free units a day; Basic $9 per month (as of 2026-10)

Limitation: You cannot load your own local checkpoint files as in Invoke's Model Manager; personalization is through LoRA training on up to 50 images on most plans.

Visit Krea

3. ImageQuarry

Best for: Changing an image by describing it, with no install and no layers to manage

How it differs from Invoke (InvokeAI): Invoke gives you layers, masks and control layers to steer each edit. ImageQuarry works by conversation instead: you select a version, describe the change and what to keep, and get a new version while the earlier ones stay. It runs in the ImageQuarry app, with four models to choose from and Nano Banana 2.1 as the default.

Price: From $9 a month (Starter); Pro $29 a month (as of 2026-10)

Limitation: Every plan is paid, while Invoke costs nothing, and the model choice is the four listed models rather than your own files.

4. AUTOMATIC1111

Best for: Tinkerers with older or smaller GPUs who want a free local web UI

How it differs from Invoke (InvokeAI): AUTOMATIC1111's web UI lists 4GB video card support, a lower bar than Invoke's 8GB guideline for SDXL, and installs on Nvidia, AMD, Intel and Apple Silicon. It covers txt2img, img2img, inpainting and outpainting, and saves generation parameters inside each file, much like Invoke.

Price: Free and open source (as of 2026-10)

Limitation: The last release, v1.10.1, is from February 2025, and it names no instruction or conversational editing feature.

Visit AUTOMATIC1111

5. Fooocus

Best for: Beginners who want the simplest free local SDXL generator

How it differs from Invoke (InvokeAI): Fooocus strips image generation down to a prompt box, with inpaint, outpaint and variations under Input Image, and runs offline on a 4GB Nvidia GPU with 8GB of RAM. It is the easiest local option here, far simpler than Invoke's canvas and nodes.

Price: Free (non-commercial open source) (as of 2026-10)

Limitation: It is in bug-fix-only long-term support with no plans for newer model architectures, and its README recommends other tools for Flux.

Visit Fooocus

Who should stay with Invoke (InvokeAI)

You have a capable Nvidia GPU or an Apple Silicon Mac and want a free, open-source tool you control completely.
You rely on the layered canvas, regional references and inpaint masks for precise control over each part of an image.
You train or collect your own LoRAs and checkpoints and want them in one Model Manager.
You share one machine with a few people on a local network and Multi-User mode covers it.

How to move from Invoke (InvokeAI)

Use Download Board Images to package each board as a.zip, and download videos one by one, since the board export leaves them out.
Keep the original files: Invoke stores the prompt, seed and models inside each image, so you can read your settings back later.
Save any Canvas Project you may reopen as an.invk file before changing tools.
Recreate one recent piece in the alternative from its prompt and reference images, and compare the result with your Invoke version.
Delete boards only after the downloads finish, choosing Delete Board Only if you want to keep the images.

Does not carry over: Canvas Projects (.invk files), node workflows and Model Manager setups are Invoke formats and do not open in other tools.; External model keys in api_keys.yaml belong to your provider accounts and must be set up again elsewhere.

Invoke (InvokeAI) alternatives: questions

Is Invoke still available?

Yes, as free open-source software you install yourself, with v7 alpha releases in October 2026. The hosted Invoke.ai platform has shut down, and invoke.com is a parked domain, not the product.

What is a hosted alternative to Invoke?

Comfy Cloud is the paid hosted version of ComfyUI, from USD 20 a month billed monthly, and Krea runs fully in the browser with a free plan of 100 units a day. ImageQuarry also runs in its own app, and every plan is paid.

Which Invoke alternative runs on a weaker GPU?

AUTOMATIC1111 lists 4GB video card support, Fooocus needs 4GB of Nvidia VRAM and 8GB of RAM, and ComfyUI says it can run large models on as little as 4GB VRAM. Invoke's guideline for SDXL is 8GB VRAM and 16GB RAM.

Invoke or ComfyUI?

Both are open source with node workflows. Invoke centers on a layered canvas with boards, while ComfyUI centers on the node graph, adds a paid cloud and a Comfy Agent, and lists far more model families.

How is ImageQuarry different from Invoke?

Invoke is software you run and tune yourself, with layers, masks and your own models. ImageQuarry is a workspace where you ask for each change in words, keep every version and branch from earlier ones, and organize work in projects.

Sources

Checked October 20, 2026. Recheck public pages before making a buying decision.

Checked October 20, 2026

Krea alternativesAlternatives to Krea for phone apps, credits that last and a simpler editing loop.Read more Stable Diffusion alternativesAlternatives to Stable Diffusion for local interfaces, newer open models and a hosted app.Read more Civitai alternativesAlternatives to Civitai for keeping your generations, inpainting and running LoRAs elsewhere.Read more All AI image tool alternativesEvery alternatives page, grouped by kind of tool.Read more

Create an image. Edit it with words. Keep refining.

Skip the setup and ask for the change

Describe an image or upload one, ask for changes in plain language, and keep every version in a project you can come back to.

Start creating Already have an account? Sign in