testingcatalog.com
|
ksl
|
|
A UI leak from Gemini’s video generation interface revealed a reference to “Omni,” a new model that may unify image and video generation into a single system. Currently Gemini uses Veo 3.1 for video and Nano Banana for images as separate tracks – Omni would consolidate them, potentially making Gemini the first major multimodal model with native video output rather than a bolted-on generation tool. The timing points toward a Google I/O 2026 announcement on May 19-20. ByteDance’s Seedance 2.0 currently leads video generation benchmarks, which likely explains the urgency. Google moving toward a unified omni-model architecture for media generation would be a meaningful structural shift away from the modular approach most labs still use.
