Amuse 2.0 beta released for easy on-device AI image generation on modern AMD hardware

AMD
(Image credit: AMD)

We all use artificial intelligence one way or another these days, it may be built-into Microsoft's Word or work behind Google's Gmail. But a major difference between the former and the latter is that the Microsoft app does not need an Internet connection, while Gmail does. The same thing applies to OpenAI's ChatGPT's image generation: it needs a connection. But how about a program that does not? Welcome, AMD Amuse 2.0.0 Beta.

Amuse 2.0.0 Beta is a program meant to enable users generate high-quality images with AI models on modern AMD hardware-based PCs. AMD's Amuse requires AMD Ryzen AI 300-series processors with at least 24GB of RAM, or a Radeon RX 7000 system with at least 32GB of memory. It will also work on systems built around Ryzen 8040 series processors, as long as they have at least 32GB of RAM. So we are talking about quite modern and fairly premium systems. 

(Image credit: AMD)

Let's talk about what it can do in more detail. AMD's Amuse 2.0.0 Beta enables users to generate high-quality images with AI models, convert paintings and drawings into digital images, and create custom AI filters.

One of the key features of the AMD Muse is the usage of AMD XDNA Super Resolution, which upscales images from 512x512 to 1024x1024 - speeding image generation times. AMD says users can create images from scratch, convert their hand-drawn artwork into digital format, and apply or create custom AI filters to modify the appearance of existing images.  

But while AMD's program definitely enables new new AI-generated content, publishers should probably beware of accidental copyright infringement. Typically, LLMs are trained on vast datasets that often include copyrighted material, without explicit permission from the content creators. This can lead to issues on platforms like YouTube due to copyright strikes - where images used are a close enough match to copyrighted content.

Additionally, there is a risk of the models generating content that contains substantial portions of copyrighted text, raising concerns about unauthorized use and distribution. The lack of Internet connection to detect training data sources could also be an issue.

AMD's Amuse 2.0 Beta is available now via the source link, top. It is a download of just over 800MB. If some app functionality doesn't work as expected, please remember that this is a beta release and AMD explains it is "experimental software."

Anton Shilov
Contributing Writer

Anton Shilov is a contributing writer at Tom’s Hardware. Over the past couple of decades, he has covered everything from CPUs and GPUs to supercomputers and from modern process technologies and latest fab tools to high-tech industry trends.

  • Makaveli
    The timing of this is good I just started using LLM's on my Radeon's and now Amuse 2.0.0 beta is out. Downloading now.

    on reddit people are saying this works on 6000 series also.
    Reply
  • brandonjclark
    Makaveli said:
    The timing of this is good I just started using LLM's on my Radeon's and now Amuse 2.0.0 beta is out. Downloading now.
    If you haven't been introduced to the world of models and datasets just yet, have fun exploring the nerds playground at HuggingFace.co.
    Reply
  • Makaveli
    brandonjclark said:
    If you haven't been introduced to the world of models and datasets just yet, have fun exploring the nerds playground at HuggingFace.co.
    Actually the other wccftech article about LM studio getting Rocm support got me in a few days ago.

    lol I spent most of the weekend messing with LLM studio and different models.
    Reply
  • NinoPino
    Makaveli said:
    The timing of this is good I just started using LLM's on my Radeon's and now Amuse 2.0.0 beta is out. Downloading now.

    on reddit people are saying this works on 6000 series also.
    How it works ? Works well ?
    Reply
  • Makaveli
    NinoPino said:
    How it works ? Works well ?
    If you have RDNA 2 download it and try.
    Reply
  • derekullo
    If the app+model is only 800 megabytes in size I don't think the quality will be that good, unless the 800 megabytes was just for the software without the model.
    Reply
  • Persister
    " can convert paintings and drawings into digital images"
    isn't that what scanners have been doing for 25 yrs.?
    Reply
  • lightofhonor
    derekullo said:
    If the app+model is only 800 megabytes in size I don't think the quality will be that good, unless the 800 megabytes was just for the software without the model.
    Model not included. Looks like it uses SDXL Turbo on Fast, SDXL Base on Balanced, and Juggernaut XL on Quality. Each is about 12GB, at least according to the download. All the other functions also need a new model as well.

    Also confirms with works on my Radeon VII, though XDNA Super Resolution does not work for me.
    Reply
  • Makaveli
    lightofhonor said:
    Model not included. Looks like it uses SDXL Turbo on Fast, SDXL Base on Balanced, and Juggernaut XL on Quality. Each is about 12GB, at least according to the download. All the other functions also need a new model as well.

    Also confirms with works on my Radeon VII, though XDNA Super Resolution does not work for me.
    This!!

    XDNA SR is only for NPU's not available on my 7900XTX.
    Reply
  • NinoPino
    Admin said:
    Amuse 2.0 beta released for easy on-device AI image generation on modern AMD hardware : Read more
    The system requirements in the article are recommended and not required.
    Reply