/>
X
Innovation

AI art generator: Qualcomm gets Stable Diffusion running on a smartphone

As smartphones become increasingly capable, more workloads are moving on-device and off the cloud. Qualcomm's MWC demo showcases the work of its AI research team.
Written by Charles McLellan, Senior Editor

AI art generators are fun and increasingly popular tools for creating images and artworks by typing a prompt and allowing the AI model, which will have been trained on a dataset of billions of images, to do its stuff. OpenAI's DALL-E 2 tool is the leading example, but there are several others, including Stability AI's Stable Diffusion.

Stable Diffusion is an open-source large language AI model trained on a subset of 2.3 billion English language-labeled images from the LAION-5B dataset, which contains 5.85 billion image-text pairs. You can access Stable Diffusion running in the cloud via the web, or run it on a PC with a GPU packing 6GB or more of video RAM, or on an Apple Silicon device running MacOS (13.1) or iOS (16.2). The latest version of Stable Diffusion, 2.1, was released in December 2022. 

Also: These experts are racing to protect AI from hackers. Time is running out

And now at Mobile World Congress (MWC), Qualcomm will for the first time demonstrate Stable Diffusion running locally on a premium smartphone powered by the company's latest Snapdragon 8 Gen 2 chipset. 

Qualcomm AI Stack diagram
Qualcomm

This development is the work of the Qualcomm AI Research team, leveraging Qualcomm's AI Stack and developer tools. The Stable Diffusion language model is quantized and optimized using int8, which allows the entire model to be stored on the smartphone. Four AI stack optimizations can generate a high-quality, 512-by-512 pixel image in under 15 seconds from the time of query entry, showing 20 inference steps going from random pixels to the finished image. "We are continually improving latency times and accuracy," said Mike Roberts, VP, head of global product, partner, and technology marketing at Qualcomm, at a pre-MWC briefing. 

The move towards processing AI workloads on-device rather than in the cloud is driven by performance, privacy, security, and cost considerations. "The more that can be on the device, at the edge, the more scalable we can make this technology across applications," said Roberts.

"We really want to showcase that this is how we're going to be able to scale across devices and foundation models to make AI truly ubiquitous," he added.

Also: What is ChatGPT? Here's everything you need to know

Here are some examples of what Stable Diffusion can do (on the web):  

Stable Diffusion rendering of Picasso-style guitar
Stable Diffusion
Stable Diffusion rendering of lonely as a cloud
Stable Diffusion

A cautionary note: The huge databases used to train AI art generators scrape images from the internet and might include inappropriate or copyrighted material. Getty Images is currently suing Stable AI for allegedly using 12 million of its images from the LAION-5B dataset without permission to train Stable Diffusion.

Editorial standards