AI videos and images for everyone: Amuse 3.0.1 simplifies the use of AI

The free Windows program Amuse has been released in version 3.0.1. It offers easy access to image generation with artificial intelligence – on the local computer, without internet access.

The software is sponsored by AMD and therefore has some optimizations for Ryzen CPUs with AI coprocessors, but also runs on Nvidia graphics cards, for example. If you lack accelerator hardware, Amuse also enables rendering on the CPU; however, this takes considerably longer.

Local BiId and video AI thanks to Amuse

The software can be downloaded directly from the Amuse project website. The recently released version 3.0.1 uninstalls any previously installed older versions and the AI models downloaded with them, as they are no longer compatible. Some new stable diffusion models are available instead. The tab for creating videos had already been created and could be used to modify existing videos – but now eight new models have been added with "Locomotion", which support "Text to Video", i.e. the creation of video clips based on text prompts.

This means that a very young branch of generative AI has now also arrived in the free sector. Various aspect ratios are preset, but different resolutions including "HQ" (for high quality) can be set. By default, Amuse generates 2-second video snippets at 24 frames per second, but this can also be configured further.

In initial tests, a two-second clip of 640×384 pixels in size on a notebook with Ryzen 7640HS Pro with 64 GB RAM and a fast PCIe Gen4 SSD takes around two and a half minutes to process. Longer clips and higher resolutions require significantly more time. It also takes a lot of finesse to develop prompts that produce the desired result. Some models also have inherent problems with certain representations – instead of a dog's face in the example image above, they generate the faces of Asian people. Movements also often look unnatural.

Amuse invites you to try out creating your own AI images and videos with the different models. Other functions include upscaling with AI or "feature extraction" from images, which can then also be used to create new image content using artificial intelligence.

Version 3.0.1 of Amuse provides more models for download, including several that are optimized for AMD processors. The downloading of models has been improved: the individual selected models are placed in a download queue and downloads can be resumed after they have been canceled. If the models are on the computer, Amuse no longer requires an Internet connection. The generated content and prompts are therefore stored on the user's own computer and not on external servers –, which means that the AI can be used without compromising privacy.

(dmk)