We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control
deepmind.google·6d ago
AI Engineering
DeepMind Blog·May 17·5 min read
Introducing Gemini Omni
TL;DR
Previous AI models could generate images but lacked video creation and editing capabilities grounded in real-world knowledge. DeepMind built Gemini Omni, a multimodal model accepting images, audio, video, and text as input to generate and edit high-quality videos through conversational commands.
✦ Why It Matters
Engineers can now build video creation and editing features into applications using conversational interfaces instead of complex traditional editing tools.
Key Takeaways
- Experiment with video editing using natural language commands.
- Combine multiple media types for richer content creation.
- Utilize Omni's physics understanding for realistic scenes.
- Access Gemini Omni Flash through Google AI and YouTube Shorts.
How It Works
Gemini Omni Flash leverages advanced natural language processing to interpret user commands, allowing for seamless video editing and creation. It maintains contextual continuity across edits, ensuring that changes build on previous instructions.
The model's understanding of physics enables it to generate realistic movements and interactions within the video, enhancing the storytelling aspect.