Snap has unveiled new augmented reality (AR) experiences powered by generative AI (GenAI) models. The company highlights breakthroughs that optimize GenAI models to run in real-time on mobile devices.

Snap previews a real-time image model that instantly generates vivid AR experiences from text prompts. This early prototype leverages faster and lighter GenAI techniques developed by Snap’s research team to enable on-device AI model execution.

The optimized GenAI models already power several Snapchat features, such as Bitmoji Backgrounds, customizable Chat Wallpapers, AI “Dreams” filters, virtual AI Pets, and AI Lenses that apply visual effects to snaps.

For AR creators, Snap has introduced a new GenAI Suite in its Lens Studio AR authoring tool. This suite of tools generates custom machine-learning models and assets, accelerating lens development timelines from weeks or months to significantly faster turnarounds.

Snap partners with London’s National Portrait Gallery to create AI-generated lenses inspired by iconic portrait styles using the GenAI Suite. Users can apply these effects to selfies and submit their “Living Portrait” to a gallery exhibition.

The GenAI Suite launched as part of the new Lens Studio 5.0 update, which is focused on productivity, modularity, and performance enhancements. The release provides expanded tools for personalizing development workflows and building more complex AR projects.

Snap states that the GenAI Suite empowers artists, creators, and developers to unlock new creative dimensions by combining custom AI models with additional AR capabilities in Lens Studio.