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·Oct 25·5 min read
Behind “ANCESTRA”: combining Veo with live-action filmmaking
TL;DR
Filmmakers faced challenges in blending live-action with generated video content. Google DeepMind developed Veo, a video generation model, enhancing it for personalization and motion matching.
✦ Why It Matters
Engineers and researchers can explore AI-driven content generation to enhance creative processes in filmmaking and other media.
Key Takeaways
- Utilize Veo for generating complex scenes in your projects.
- Incorporate AI-generated visuals to enhance storytelling.
- Experiment with Gemini for detailed prompt creation.
- Leverage new capabilities for blending live-action and AI footage.
How It Works
The film 'ANCESTRA' uses Veo to generate video content that matches the quality and motion of live-action scenes. By fine-tuning the Imagen model, the team ensured consistent art direction across AI-generated elements.
They also created a virtual 3D model to guide camera movements, allowing for precise motion matching in generated videos.