Amuse AI: First Updates Following Open Source Transition

Amuse, the local AI image generation software developed by TensorStack, released version 3.2.0 on April 2, 2026. The update signals a return to active development for the tool after reports in December 2025 suggested the project had been abandoned and its source code open-sourced.

According to reporting from heise online, the release of Amuse 3.2.0 indicates that the local image AI is continuing to evolve despite previous indications that development had ceased.

Technical Infrastructure Upgrades

The v3.2.0 release implements a significant migration of the application’s entire stack. The software has moved to the .NET 10 framework, which TensorStack states provides improved memory management and the latest Just-In-Time (JIT) optimizations for high-performance C# applications.

The core inference engine has been updated to OnnxRuntime v1.23.0. This update is intended to provide faster generation speeds and broader support for current model architectures. All internal NuGet libraries have been updated to their latest stable versions to improve security and stability.

Amuse is built upon the OnnxStack engine. While the compiled application is distributed for users, the core logic of both Amuse and OnnxStack remains available to the community under the Apache 2.0 license.

Development Context and Open Source Transition

The latest update follows a period of uncertainty regarding the project’s status. On December 10, 2025, reports from Videocardz and discussions on Reddit noted that Amuse had gone open source, with some observers suggesting the move occurred because the project had been effectively ended months prior.

The transition to an open-source model allowed the software to work with all GPU vendors, moving beyond its initial tight integration with specific hardware partners.

Hardware Optimization and Compatibility

Amuse is designed to run on Windows PCs, utilizing the computer’s GPU, CPU, or NPU power to generate images from prompts without requiring a login, account, or internet-based storage for prompts and images.

The software is highly optimized for AMD hardware through a partnership between TensorStack and AMD. Certain features are exclusively available for systems equipped with AMD XDNA™.

Recommended specifications for AMD hardware include:

• AMD Ryzen™ AI 300 series processors with 24GB of RAM or higher.
• AMD Ryzen™ 8040 series processors with 32GB of RAM, noting that AMD XDNA™ Super Resolution requires the latest NPU driver and OEM MCDM updates.
• Radeon™ RX 7000 series graphics cards.

AI Capabilities

The software allows users to create AI art using Stable Diffusion models and newer AI models such as Flux. Users can generate images from text prompts, apply custom AI filters to existing photos, and use AI collaboration tools to draw new art.

The application includes a model management system that allows users to select and manage the specific AI models they use to match their preferred art styles and hardware capabilities.