Skip to content

How-ToGPU

How to Generate AI Images Locally on Windows with Amuse

How to make AI images locally on Windows with Amuse, the free open-source studio that runs FLUX, SDXL, Qwen and video models on NVIDIA cards through CUDA or on AMD, NVIDIA and Intel graphics through Vulkan. What changed since its AMD days and how to install it safely.

By , Editorial team

4 min read

How to Generate AI Images Locally on Windows with Amuse

If you want to make AI images on your own PC without wrestling with Python, command lines or node graphs, Amuse is one of the simplest options on Windows. It is a free desktop app that runs image, video, speech and text models on your own graphics card, so your prompts and pictures never leave your PC.

Amuse has changed a lot since it was launched as an AMD-backed app, and many older tutorials are now out of date. This guide explains what Amuse is today, what hardware it needs and how to install it from the right place. Forge PC has not benchmarked Amuse, and we do not quote generation times, because speed depends on your graphics card, driver, model and settings.

What Amuse is now

Amuse began as a collaboration between TensorStack, a New Zealand software company, and AMD. According to ModelTrends' history of the project, Amuse 3.0 added AMD-optimised models in 2025, and in December 2025 Amuse was released as open source. ModelTrends reports that the community-maintained versions since then have dropped some AMD-specific extras, such as the XDNA Super Resolution upscaler, and added support for other GPU brands.

TensorStack's current Amuse page describes it as "an open-source creative studio for Windows" and the reference application for its TensorStack SDK. The latest release listed there is v3.8.7, released on 29 October 2026, built on .NET 10, with a 407MB installer.

The licence is worth understanding: TensorStack says the source code is Apache-2.0, while the compiled app is distributed under the Amuse 2.0 License, because it bundles third-party software with its own terms. The app is free to use.

What it can do

TensorStack lists features including image generation and editing, inpainting, LoRA and ControlNet support, image and video upscaling, frame interpolation, a video editor, speech recognition with Whisper, voice generation and local text generation with Qwen and Gemma models. It loads models in Safetensors, GGUF and ONNX formats.

Supported image models include Z-Image, Qwen, FLUX.1, FLUX.2, Stable Diffusion XL and Stable Diffusion 3, among many others. Video models include Wan 2.2 and CogVideoX. New pipelines arrive often, so check the release notes on GitHub for the latest list.

Behind the scenes, Amuse uses three engines: stable-diffusion.cpp for image and video diffusion, Hugging Face Diffusers and Transformers pipelines running in managed Python environments, and ONNX Runtime for language, speech and vision tasks such as upscaling. You do not need to set any of these up yourself.

What hardware do you need?

Amuse offers two GPU backends:

Backend

Works with

Requirements

CUDA 13.0

NVIDIA graphics cards

NVIDIA driver 580.65 or later. TensorStack strongly recommends RTX cards; older Pascal and Maxwell cards may work but are not recommended.

Vulkan

AMD, NVIDIA and Intel GPUs

Support depends on your GPU and its Vulkan driver. No CUDA or ROCm needed.

TensorStack warns that real-world compatibility and speed depend on your GPU, driver, operating system and backend. Video memory is the main limit on which models you can run comfortably. Our guide to building a PC for Stable Diffusion and FLUX explains how much VRAM different models need.

How to install Amuse

  1. Update your graphics driver. NVIDIA users need driver 580.65 or newer for the CUDA backend. Our driver update guide covers NVIDIA, AMD and Intel.

  2. Download from the official GitHub Releases page for saddam213/AmuseAI, or via TensorStack's Amuse page, which links there. ModelTrends warns that old download links can point to discontinued versions, so avoid installers from third-party download sites and old tutorials.

  3. Choose the installer or the .zip archive and install it like any Windows app.

  4. If setup fails on Windows 11, TensorStack says to run Amuse as administrator so it can set up its environment.

  5. Pick your backend: CUDA on an NVIDIA RTX card, or Vulkan on AMD and Intel graphics.

Tips for your first images

  • Start with smaller models. ModelTrends says early community builds had stability problems with bigger models. Even though later releases have moved on, a smaller model such as Stable Diffusion XL is a safe first test before you try large models like FLUX.

  • Lower the resolution or use a smaller or quantised model if you get a graphics card error or run out of memory.

  • Close other GPU-heavy apps, such as games, while generating.

  • Keep an eye on temperatures during long image or video jobs, using the tools in our PC temperature monitoring guide.

  • Check licences before commercial use. Each AI model has its own licence terms, separate from Amuse's.

Amuse or something else?

Amuse is the easy option: one app, no command line. If you want more control or more models, other tools may suit you better:

  • ComfyUI is far more flexible, using a node-based workflow. See our ComfyUI install guide.

  • Stability Matrix installs and manages several image-generation apps for you. See our Stability Matrix guide.

  • On a Radeon card, ModelTrends suggests ComfyUI with official ROCm support for serious work on recent cards. Our AMD local AI guide covers the options.

  • To enlarge existing photos, a dedicated upscaler like Upscayl is simpler still.

Our verdict

Amuse is a good way to start making AI images, video and speech locally on Windows without any coding. It is free, open source, works on NVIDIA cards through CUDA and on AMD, NVIDIA and Intel graphics through Vulkan, and supports a wide range of current models. Download it only from the official GitHub releases, start with small models and keep your graphics driver current. If you outgrow it, ComfyUI offers more control. Forge PC has not benchmarked Amuse; this guide is based on TensorStack's documentation, the project's GitHub releases and ModelTrends' history of the app.

The bottom line

Amuse is a good way to start making AI images, video and speech locally on Windows without any coding. It is free, open source, works on NVIDIA cards through CUDA and on AMD, NVIDIA and Intel graphics through Vulkan, and supports a wide range of current models. Download it only from the official GitHub releases, start with small models and keep your graphics driver current. If you outgrow it, ComfyUI offers more control. Forge PC has not benchmarked Amuse; this guide is based on TensorStack's documentation, the project's GitHub releases and ModelTrends' history of the app.

Sources

  1. 01TensorStack
  2. 02GitHub
  3. 03ModelTrends

About the author

Forge PC Staff

Editorial team

The shared byline for articles researched, written and edited by the Forge PC editorial team. How we source, check and correct our work is set out in our editorial policy.

More from Forge PC Staff

Read next

How to Enable AMD Fluid Motion Frames (AFMF 2.1) on a Radeon Graphics Card

How-To

How to Enable AMD Fluid Motion Frames (AFMF 2.1) on a Radeon Graphics Card

To enable AMD Fluid Motion Frames, install Adrenalin 25.3.1 or newer, then select the HYPR-RX profile (RX 7000 and newer) or turn on AMD Fluid Motion Frames under Gaming > Graphics, globally or per game. AFMF 2.1 works on RX 6000 cards and newer in DirectX 11, DirectX 12, Vulkan and OpenGL games, with VSync off; RX 6000 cards need exclusive full screen.

Forge PC Staff2 min

AMD's FSR 4.1.2 Unreal Engine 5 Plugin Brings ML Upscaling to RX 6000 Cards

News

AMD's FSR 4.1.2 Unreal Engine 5 Plugin Brings ML Upscaling to RX 6000 Cards

AMD's FSR Unreal Engine 5 plugin now uses FSR Upscaling 4.1.2, bringing ML-based upscaling to Radeon RX 6000 cards and select Ryzen integrated graphics as well as RX 7000 and RX 9000, in UE 5.3 to 5.8 games that adopt it. It's DirectX 12 only for now and adds Frame Generation 4.0.2 with driver-assisted frame pacing on RX 9000 cards.

Forge PC Staff2 min

NVIDIA's DGX Spark HPC Developer Kit Links Four Sparks Into a 512GB Desk-Side Cluster

News

NVIDIA's DGX Spark HPC Developer Kit Links Four Sparks Into a 512GB Desk-Side Cluster

NVIDIA's DGX Spark HPC Developer Kit, announced at SC26 on 16 November 2026, links four DGX Spark systems in a ring over ConnectX-7 with no network switch, for 512GB of combined memory and 20TB of NVMe storage. NVIDIA says it can spread models of up to 700B parameters across the nodes with NVFP4. No UK kit price yet; Scan sells two-Spark bundles at about £12,845.

Forge PC Staff3 min

Best Free Benchmark Software for PC in 2026: CPU, GPU, SSD and AI Tests

Best Picks

Best Free Benchmark Software for PC in 2026: CPU, GPU, SSD and AI Tests

The best free PC benchmarks in 2026: Cinebench 2026 for CPUs, Unigine Superposition Basic or 3DMark's free Steel Nomad test for graphics cards, CrystalDiskMark for SSDs, Geekbench 7 for cross-platform scores, Geekbench AI for NPUs, Blender Benchmark for creators and OCCT for stability. Run each test twice and download only from official sites.

Forge PC Staff5 min