Making a generative AI image might take less than two seconds on a smartphone in the future

There are lots of generative AI-based art creators out there, from Midjouney, to Bing Image Creator and many more. However, it sometimes takes a while for those kinds of image generators to do their job, depending on the text prompt and other factors.

Today, Snap, the parent company of Snapchat, announced that its Snap Research division has come up with a technique that should speed up the time that generative AI art creators make images.

In a blog post, Snap stated the new method is called SnapFusion, and it cuts down the time to make an image via generative AI down to less than two seconds. It claims this is "the fastest time published to date by the academic community."

The blog post added:

Snap Research achieved this breakthrough by optimizing the network architecture and denoising process, making it incredibly efficient, while maintaining image quality. So, now it’s possible to run the model to generate images based on text prompts, and get back crisp clear images in mere seconds on mobile rather than minutes or hours, as other research presents.

The research paper (in PDF format) goes into more detail on SnapFusion, along with a number of examples of artwork created by the new technique. It states:

Specifically, we propose an efficient UNet by identifying the redundancy of the original model and reducing the computation of the image decoder via data distillation. Further, we enhance the step distillation by exploring training strategies and introducing regularization from classifier-free guidance.

The bad news is that Snap indicates this new method is still going to take a while before it's available to the general public. However, it adds that it "has the potential to supercharge high quality generative AI experiences on mobile in the future." It's likely that this new technique could be added to a future update of the Snapchat mobile app.