TensorStack.AI
Amuse
v3.8.6 Windows Open source

Local AI for image, video, audio and text.

Amuse is an open-source creative studio for Windows. Generate and edit images, create video, synthesize speech and write with LLMs, using the latest open models on your own GPU.

Download for Windows Visit amuse-ai.com

v3.8.6 · .exe · 407 MB Released Oct 29, 2026 Release notes GitHub

24Image pipelines
8Video pipelines
4Audio pipelines
5Text pipelines
The official Amuse website Screenshots of every studio, real outputs, the full model library and the latest releases. amuse-ai.com
Features

One app for the whole creative workflow.

Generate, refine and finish your work in one place, on your own machine.

Safetensors, GGUF & ONNX

Load models in the formats the community actually ships.

Video editor

Edit generated clips or your own footage in the built-in editor.

Image & video upscaling

Bring stills and moving media up to a higher resolution.

Feature extraction

Pull feature maps, background masks and poses from images and video.

Frame interpolation

Raise frame rates or turn any clip into smooth slow motion.

Inpainting

Remove unwanted objects or fill areas with generated content.

Advanced image editing

Selection and masking tools for precise, targeted edits.

LoRA & ControlNet

Steer style and composition with adapters and control maps.

Voice generation

Turn text into speech with Supertonic voices.

Speech recognition

Transcribe recordings and video audio with Whisper.

Text generation

Run Qwen and Gemma language models locally.

Media gallery

Organize, browse and manage everything you create.

Supported models

The open models people are actually using.

New pipelines arrive with regular releases. Check the release notes for the latest additions.

Image

24
  • Z-Image
  • Qwen
  • FLUX.1
  • FLUX.2
  • FLUX.3
  • Chroma
  • Kandinsky 5
  • Stable Diffusion XL
  • Stable Diffusion 3
  • Ernie Image
  • Anima
  • JoyAI Image
  • PRX-Pixel
  • Krea2
  • GLM Image
  • Ideogram 4
  • Boogu
  • Lens
  • LongCat
  • HiDream-O1
  • LLaDA Image
  • SenseNova U1.5
  • MOSS-Image
  • Fibo 2.0

Video

8
  • Wan 2.2
  • CogVideoX
  • Kandinsky 5
  • SkyReels-V2
  • Helios
  • Motif
  • AnyFlow
  • MiniMax-H3

Audio

4
  • ACE-Step XL
  • Whisper
  • Supertonic v3
  • LongCat Audio

Text

5
  • Qwen 3.0
  • Qwen 3.5
  • Qwen 3.6
  • Qwen 3.8
  • Gemma 4.0
Under the hood

Three inference engines, one studio.

Amuse is the reference application for the TensorStack SDK, which brings each model family to the runtime that suits it best.

stable-diffusion.cpp

Fast GGUF and Safetensors diffusion for image and video generation, with CUDA and Vulkan backends.

Hugging Face

Diffusers and Transformers pipelines running in managed Python environments, driven from .NET.

ONNX Runtime

Optimized ONNX models for language, speech recognition, voice and vision tasks such as upscaling.

GPU support

Pick the backend that fits your hardware.

CUDA

NVIDIA

Native GPU acceleration with CUDA 13.0. RTX cards are strongly recommended — their Tensor Cores are designed for AI workloads. Older Pascal and Maxwell cards may work, but aren't recommended.

Requires NVIDIA driver 580.65 or later.

Vulkan

AMDNVIDIAIntel

A cross-vendor backend for compatible GPUs, with no need for CUDA or ROCm. Useful for systems without a supported CUDA or ROCm environment.

Support depends on your GPU and its Vulkan driver.

On Windows 11, run Amuse as administrator if setting up an environment fails.

Real-world compatibility and speed depend on GPU model, driver, operating system and backend.

Download

Get Amuse for Windows.

Grab the latest release from GitHub, as a Windows installer or a .zip archive.

Version
v3.8.6 — Fibo 2.0
Released
Oct 29, 2026
Platform
Windows · built on .NET 10
Installer size
407 MB
Installer (.exe) Archive (.zip) All releases
Build from source

Prefer to compile it yourself?

Amuse is built on .NET 10 with Visual Studio.

  1. Install Visual Studio 2026 with the .NET desktop development workload.
  2. Clone the repository and open the .sln file.
  3. Let Visual Studio restore the NuGet packages and dependencies.
  4. Choose a Debug or Release configuration and build the solution (Ctrl+Shift+B).
  5. Run Amuse.exe from the build output folder.
git clone https://github.com/saddam213/AmuseAI.git
License

Open source, with bundled dependencies noted.

The Amuse source code is licensed under Apache-2.0. Compiled binaries are distributed under the Amuse 2.0 License, because they bundle third-party software that may carry its own licensing terms.

Built with

Your GPU is ready. Are you?

Download Amuse and start creating locally, or see everything it can do at amuse-ai.com.

Download Amuse Visit amuse-ai.com