Qualcomm runs AI art generator Stable Diffusion on an Android phone
Qualcomm announced that it has conducted the world's first demonstration of the artificial intelligence (AI) art generator Stable Diffusion on an Android smartphone. The smartphone used for the demo ran on Qualcomm’s latest Snapdragon 8 Gen 2 processor, which powers the new Samsung Galaxy S23 series.
Qualcomm stated in a blog post that the Stable Diffusion model, which typically requires a great deal of compute power, can now generate photos in a matter of seconds directly on a smartphone.
Stable Diffusion is a deep-learning model that can generate artwork and images from text prompts. It was developed by UK-based Stability AI and made public in August.
In a demo video, Qualcomm showed that the updated version 1.5 of Stable Diffusion can create a picture of 512x512 pixels in just 15 seconds.
According to Qualcomm, “The techniques were applied across all the component models in Stable Diffusion, namely the transformer-based text encoder, the VAE decoder, and the UNet. This was critical for the model to fit on the device.”
Further Qualcomm said that it used its AI Engine direct framework for compilation the neural network into a program that can run efficiently on a smartphone.
Qualcomm isn’t the first to try out the AI art generator on smartphones. In December, Apple released updates that will allow Stable Diffusion to run locally on iPhones using the Draw Things app with Core ML, which is Apple's proprietary framework for machine learning.
In Apple’s demo, it took about a minute to generate a 512x512 image, which is four times taken by a smartphone running on Qualcomm’s chip.
Further, Qualcomm said that the optimisations made for Stable Diffusion's efficient operation on mobile devices can also be emulated on laptops, XR headsets or any other device powered by Qualcomm.
Qualcomm claims that this would lower the prices that would otherwise be significantly higher on cloud-based alternatives. “All of this is a step in the direction of Qualcomm's goal of integrating AI improvements into its products and making those improvements more accessible to consumers.”