Soon your Android phones might be able to generate AI images within seconds

The ability for Android phones to generate AI images was demonstrated by Qualcomm. These images were created within seconds using simple text prompts. Chipmaker has created an image using text to image generative model that was 512 x 512 pixel resolution of a warrior cat with use of android phone and stable diffusion.

AI images on smartphones using text prompt

As stated by Qualcomm, its AI stack was able to run diffusion 1.5 on iPhone that wasn’t even connected to internet. All the operations that were necessary to acquire the result were performed on device itself. The video shared by company is clearly showing the generation of image within 15 seconds.

The image was created by using a text based prompt; “Super cute fluffy cat warrior in armor, photorealistic, 4K, ultra detailed, vray rendering, and unreal engine”.

The device that generated the AI image contained Qualcomm Snapdragon 8 Gen 2 that is similar Galaxy S23 series. Prior to testing, the Stable Diffusion model was tuned for “quantization, compilation, and hardware acceleration.”

It has happened before for Stable Diffusion to produce an image on an Android phone. The Verge notes that Ivon Huang, a developer, previously used Stable Diffusion on a Sony Xperia 5 II with an 8GB RAM and Qualcomm Snapdragon 865 processor. Huang, however, needed an hour to produce a 512 × 512 image. So, Qualcomm’s method is undoubtedly far quicker than the developer’s.

These kind of AI models operating on Android phones can reduce development costs, do away with the requirement for a mobile connection, and safeguard user privacy as the data is stored locally on the device.

Regarding the release date for these models on high-end Android smartphones, Qualcomm has not yet provided any information to the public. We might learn more about the technology in the next days because the company will demonstrate it at MWC 2023.