Gemini Omni: Google's Any-to-Video Multimodal Model (Google I/O)

Melvin Vivas · X video post · 2026-05-19 · 0:07 · 59 views · Open on X

Topics: Industry Trends & Job Market, LLM Fundamentals · Level: beginner

Summary

This is a short reaction post. Melvin Vivas quotes Google's Google I/O announcement of Gemini Omni, a model meant to create anything from any input, starting with video. You can mix images, audio, video and text as input to generate high-quality videos, or use drawings to guide what gets generated. The post is a trend signal about multimodal generation and has no tutorial.

Key points

• Gemini Omni was announced at Google I/O (#GoogleIO) and is described as able to 'create anything from any input'.
• Video is the first output it supports.
• Inputs can be images, audio, video and text, combined in one request, to generate high-quality videos.
• Drawings or sketches can be used to steer what gets generated so it matches your vision.
• The trend: generative models are moving from text-only to any-input, any-output multimodal systems.
• The 7-second clip has no speech, so the quoted announcement text is the only content.

Resources mentioned

• Gemini Omni · tool · deepmind.google · check price
  Google's multimodal generative model that takes combined image, audio, video, text and drawing inputs and generates high-quality video.
  Also in: Opinion: Seedance 2.0 Still Beats Gemini Omni for Video (Melvin Vivas on X · notes)
• Google I/O · other · io.google · free
  Google's yearly developer conference, where Gemini Omni was announced.