Fast AI image generation is now possible on an Android phone. In a recent demo video, Qualcomm was able to generate a 512 x 512-pixel image of a cat wearing armor using AI-image-generator Stable Diffusion version 1.5 on an Android phone. And the kicker is it finished in less than 14.42 seconds, making it pretty close to what PCs can do with the tech.
To achieve this, company engineers used a process called “quantization”. Without getting too deep in the weeds here, they basically took the desktop version of Stable Diffusion, shrunk it down, and optimized it for mobile devices. According to Qualcomm, the process “increases performance [while] also [saving] power by allowing the model” to run more efficiently thereby “[consuming] less memory bandwidth.”
It’s unknown exactly what mobile device was used in the demonstration. However, we do know it was powered by the Snapdragon 8 Gen 2 Mobile Platform, the exact same chipset found on the Samsung Galaxy S23. It is worth noting that other developers have managed to put Stable Diffusion on the Sony Xperia 5 II, but generating images on that phone took around an hour because of the weaker hardware. The main takeaway from Qualcomm’s achievement is its speed and what it could mean for users down the line.
Future of generative AI
The thing about generative AI models is they require a fair amount of computing power to work. Stable Diffusion’s developers, for example, recommend using a computer equipped with NVIDIA hardware and, at least, 6.9 GB of VRAM specifically (video RAM) to create images. Smartphones don’t come with VRAM as manufacturers opt for regular RAM to run on-device software. But now this power discrepancy is beginning to fade, so we could see this technology crop up on more mobile devices.
Jilei Hou, vice president of engineering at Qualcomm, told VentureBeat that the lessons learned in optimizing Stable Diffusion for smartphones could be utilized for other devices running on company hardware. Hou specifically points to laptops and XR headsets as potential future houses for generative AI. As for what they’ll be able to do, there are a number of applications like “image editing [plus] in-painting”, a process where an AI fills the missing parts in a generated artwork. The company also hopes all this can be done without needing an internet connection to the Stable Diffusion cloud. As cool as this may be, it’s unknown if or when this tech will see a widespread launch on phones running Qualcomm hardware.
Besides generating images, developers have created writing AI to help with writing content. Be sure to check out our recent list of the best AI writers for 2023 – Jasper ranks among our favorites.
Source: www.techradar.com