InvokeAI

InvokeAI is a free, open-source creative engine for AI-powered image generation and visual content creation. It is designed primarily for artists, designers, developers, and other creative users who want to run generative AI locally while retaining control over their models, workflows, files, and computing environment. The project is self-hosted, customizable, and currently licensed under Apache 2.0.

Unlike cloud-only AI image generators, InvokeAI can be installed directly on a user’s computer. Current downloads are available for Windows, macOS, and Linux, with Docker and manual Python installation also supported for advanced users. Windows and Linux installations support NVIDIA and AMD GPUs, while the macOS version requires Apple Silicon.

The project has evolved considerably from its earlier Stable Diffusion-focused releases. Its current release series supports a broad range of modern image-generation and editing models, including FLUX, Qwen Image, Z-Image Turbo, Krea, Ideogram, Anima, and other model families. Recent releases have also introduced video generation through Wan 2.2.

One of InvokeAI’s central features is its Unified Canvas. The canvas provides a layer-based workspace where users can generate, paint, sketch, edit, and compose images. Layers can be manipulated independently, allowing users to combine AI generation with more conventional visual editing techniques. This makes the software useful for iterative creative workflows rather than limiting users to one-shot text-to-image generation.

InvokeAI also includes a node-based workflow editor. Users can create visual graphs that connect different generation and processing operations into reproducible pipelines. These workflows can expose selected parameters through the interface, making it possible to create repeatable processes without requiring users to interact directly with the underlying graph every time.

The workflow system is particularly relevant to advanced users who want more control over their generation process. Instead of relying only on a single prompt box, users can construct pipelines involving different models, image inputs, conditioning methods, transformations, and output operations. This makes InvokeAI suitable for structured production workflows and experimentation.

InvokeAI also emphasizes rapid iteration. Its current feature set includes batch generation, prompt wildcards, and high-resolution upscaling. These features allow creators to explore multiple variations and refine concepts without constantly leaving the application.

The platform supports both locally hosted models and selected externally hosted models. For example, current documentation explains how InvokeAI can connect to OpenAI’s GPT Image family and DALL·E 3 through the OpenAI API. This allows users to combine local generation with external model providers when desired. API usage from those external providers is billed by the respective provider rather than by InvokeAI itself.

InvokeAI also provides an HTTP API for programmatic interaction. Its API supports single-user and multi-user configurations, while OpenAPI/Swagger documentation is available for the application’s endpoints. Developers can use the API to interact with workflows and build external clients or integrations around an InvokeAI installation.

The development ecosystem extends beyond image generation. InvokeAI provides documentation for developers who want to contribute to the codebase or build functionality on top of the platform. The project includes development guidance covering its architecture, frontend, testing, model integrations, and contribution process.

Another notable feature is Invoke Training, a dedicated interface for training models using techniques such as Textual Inversion and LoRA. This allows users to train models around their own styles or concepts rather than relying entirely on pre-existing models.

The project is community-driven. The current Invoke website explains that the former officially hosted Invoke platform was shut down after the founding team joined Adobe. The open-source project continues under community stewardship, with Lincoln Stein and Vic identified as core maintainers. This means the current InvokeAI product should be understood as a community-maintained open-source project rather than the former commercial hosted service.

InvokeAI has continued receiving regular releases in 2026. Version 6.14.0 added video generation using Wan 2.2, support for Krea.2, FLUX.2 Dev, Ernie Turbo, Ideogram 4, Anima ControlNets and inpainting, FLUX.2 PiD super-resolution, multi-GPU support, and native Intel XPU support. Version 6.14.1 followed with additional fixes and improvements.

The software’s hardware flexibility is another important characteristic. Current documentation supports different GPU and acceleration technologies depending on platform and configuration, including NVIDIA CUDA, AMD ROCm, Apple MPS, DirectML, Intel XPU, and CPU-based operation. The exact performance and model compatibility depend on the hardware and model being used.

InvokeAI is also available through hosted GPU services for users who do not want to run the software entirely on their own computer. The official download page currently identifies AI Badgr, RunPod, and Railway as hosted options. These services are separate from the free self-hosted InvokeAI application.

For organizations and developers, InvokeAI’s open architecture can be useful because it provides greater control over deployment and data than a conventional cloud-only image generator. However, self-hosting also means that users are responsible for their own hardware, storage, model management, updates, and security.

Security maintenance is therefore important. In 2026, the project issued security releases addressing vulnerabilities affecting server instances, including versions 6.13.7 and 6.13.8. Users operating InvokeAI as a server should keep the software updated and follow the project’s security advisories.

Overall, InvokeAI is best understood as a professional, customizable local AI creative environment rather than simply an AI image generator. Its combination of image generation, layer-based editing, node workflows, model support, training tools, APIs, and video generation gives users a broad environment for developing and producing AI-generated visual media.

Pricing

InvokeAI is free and open source. The current project does not charge a subscription for the core self-hosted application. It is licensed under Apache 2.0.

Users may still incur costs for hardware, electricity, storage, or third-party hosted GPU services. External AI providers such as OpenAI can also charge separately when their models are accessed through InvokeAI.

Review

InvokeAI is particularly suited to users who want control over their AI image-generation environment. Its layer-based canvas, node workflows, broad model support, local execution, API, and extensibility make it useful for serious creative workflows.

Its main trade-off is that self-hosting requires more technical involvement than a conventional browser-based AI generator. Users need compatible hardware or a hosted GPU provider, must manage models and storage, and should keep the installation updated.

The software’s active development is a notable feature. Releases during 2026 have expanded its model support and introduced video generation, multi-GPU support, additional hardware acceleration, and new creative tools.