InvokeAI review

InvokeAI is a self-hosted, open-source application for generating and editing images with AI. The Community Edition runs on your own hardware, so prompts, data and generated images stay local. It is free and licensed under Apache 2.0.

The main draw is its layer-based canvas: you can draw, paint and edit parts of an image without treating every change as a new generation. For repeatable or more technical work, visual node workflows let you build pipelines and expose custom parameters. InvokeAI supports Flux, SDXL, SD 1.5, checkpoints, LoRAs, Textual Inversions and ControlNets. ControlNet can use inputs such as depth maps, edges and poses to guide composition. Other workflow tools include batch generation, prompt wildcards and high-resolution upscaling.

In short, InvokeAI is a local image tool with both a visual editing interface and a node-workflow option—not a hosted image-generation service you can count on.

InvokeAI makes the most sense for people who want to generate an image and then keep editing it in the same application. Its layer-based canvas supports drawing, painting and inpainting, while ControlNet inputs can help guide composition instead of relying on prompts alone. That gives local users more direct ways to correct an image than simply generating another version.

The local setup also avoids sending prompts and assets to a hosted service. The Community Edition lists unlimited local generation and self-managed local seats; the trade-off is that you supply compatible hardware and manage the installation yourself. Model support covers several common Stable Diffusion families and related assets, including Flux, SDXL, SD 1.5, LoRAs and ControlNets.

There is a path from straightforward canvas editing to more involved automation: users can build visual, reproducible node workflows and expose parameters for reuse. Batch generation, wildcards and upscaling help with iteration. One Reddit user described installing InvokeAI, pointing it at existing models and getting started; other commenters praised the canvas and inpainting. Those are individual reports, not a guarantee that every setup will be simple.

The biggest caveat is continuity. Reddit users discussed the departure of Invoke’s commercial team and uncertainty about development; a Community Edition maintainer said the project was rebuilding its team and had a prerelease available for testing. That points to ongoing community work, but it is not the same assurance as a dependable commercial support operation. We would not choose InvokeAI if hosted service or support is a requirement.

The pricing information says the hosted tiers have since been sunset, leaving the free Community Edition active. It lists Starter at $19 monthly or $15 per month annually, with 7,200 monthly credits and 20GB storage; Indie at $49 or $39, with 18,000 credits and 100GB; and Premier at $99 or $79 per user, with 36,000 credits per user and 200GB per user. These are not current hosted options if the tiers are sunset. Enterprise pricing is custom, and the Starter trial duration is not published. For Community Edition, the practical limit is your compatible hardware, which you must provide and manage.

The published pricing snapshot lists Community Edition at $0 for unlimited local generation, with self-managed seats. The hosted plans are listed at $19 monthly or $15 per month billed annually for Starter, $49 or $39 for Indie, and $99 or $79 per user for Premier. Enterprise pricing is custom. But there’s a catch: the snapshot says the hosted tiers have since been sunset, leaving Community Edition active. Treat those paid prices as a record of listed plans, not confirmation that you can buy them now.

The listed plan limits were 7,200 monthly credits and 20GB storage on Starter; 18,000 credits and 100GB on Indie; and 36,000 credits per user and 200GB per user on Premier. That’s a jump from $19 to $49 monthly between Starter and Indie, mainly adding training and storage. Starter overage was listed at $5 per hour; Premier Flux for Commercial Use added $30 per user. A Starter trial was listed, but its duration wasn’t published. Annual prices are monthly equivalents, not annual totals.

Switching to Community Edition means running the software on compatible local hardware; InvokeAI’s product description says prompts, data, and creations stay local. It supports Flux, SDXL, SD 1.5, checkpoints, LoRAs, Textual Inversions, and ControlNets, plus layer-based editing and node workflows. If you rely on another tool’s setup, budget time to recreate workflows and check model compatibility rather than assuming they transfer. The published material doesn’t publish a migration guide or specific export formats, so verify how to move your assets before you switch.

Switching away may be more attractive if you need hosted workspaces or dependable support: our verdict flags both as poor fits, and the pricing snapshot says hosted tiers have been sunset. We weren’t given Reddit posts or comments, so we can’t attribute reasons for leaving to Reddit users. The practical trade-off in the available information is local control and free software versus hardware requirements and no confirmed hosted option.