Subvolts take Google DeepMind's Veo 3 video generation journey from research to launch is detailed. Oct 15, 2025 - Developers
- Learn about the technical hurdles and scaling challenges in building advanced video generation models like Veo 3.
- Business
- Understand how user feedback and new prompting methods are shaping the future of AI-powered video creation.
- Everyday users
- Discover the journey behind AI that can create videos, and how it is evolving to be more controlled and coherent.
This episode features a deep dive into the development of Google's state-of-the-art Veo 3 video generation model.
How a Moonshot Led to Google DeepMind's Veo 3
Audio streams from Google AI. Your place is remembered on this device.
From Subvolts
Summary and analysis by Subvolts editorial, AI-assisted.
Logan Kilpatrick hosts Dumi Erhan on Google AI: Release Notes to discuss the evolution of Google DeepMind's generative video models. They trace the journey from 2018 research to the Veo 3 model, covering technical challenges in evaluation, scaling, long-duration coherence, and how user feedback shapes AI video creation.
This episode explains how Google built its advanced AI for making videos from text. It covers the difficult steps involved, like making videos long and clear, and how people using the AI help make it better.
This episode offers insight into the development of advanced AI video generation, a key area for Google. It's for AI researchers, developers, and product managers interested in the technical hurdles and future direction of video synthesis. The discussion on scaling, coherence, and user feedback connects directly to current efforts by Google and other frontier labs to make AI tools more capable and practical for creators.
- HostLogan Kilpatrick
- GuestDumi Erhanco-lead of the Veo project, Google DeepMind
- Guest: Dumi Erhan, co-lead of the Veo project at Google DeepMind
- Show: Google AI: Release Notes
- Length: 48 minutes
- Veo Project Start: 2018
- Model Launched: Veo 3
Dumi Erhan, co-lead of the Veo project at Google DeepMind, joins host Logan Kilpatrick for a deep dive into the evolution of generative video models. They discuss the journey from early research in 2018 to the launch of state-of-the-art Veo 3 model with native audio generation. Learn about the technical hurdles in evaluating and scaling video models, the challenges of long-duration video coherence and how user feedback is shaping the future of AI-powered video creation.
Chapter:
0:00 - Intro
0:47 - Veo project's beginnings
3:02 - Veo's origins in Google Brain
5:07 - Video prediction and robotics applications
7:45 - Early progress and evaluation challenges
10:30 - Physics-based evaluations and their limitations
12:18 - The launch of the original Veo model
14:06 - Scaling challenges for video models
16:02 - The leap from Veo1 to Veo2
19:40 - Veo 3’s viral audio moment
21:17 - User trends shaping Veo's roadmap
23:49 - Image-to-video vs. text-to-video complexity
26:00 - New prompting methods and user control
27:55 - Coherence in long video generation
31:03 - Genie 3 and world models
35:54 - The steerability challenge
41:59 - Capability transfer and image data's role
47:25 - Closing
- generative video models
- Dumi Erhan
- AI systems that create videos from text or other inputs.
- coherence
- Dumi Erhan
- How well the elements in a video stay consistent and make sense over time.
- world models
- Dumi Erhan
- AI systems that try to understand and simulate the real world.
27 minKoray Kavukcuoglu on frontier models, coding agents, and building AGI
WithKoray Kavukcuoglu
39 minIntroducing Gemini Robotics 2
WithCarolina Parada, Stuart Bowers, Kanishka Rao, Jie Tan
42 minGemini co-leads on project origins and what's next
WithJeff Dean, Koray Kavukcuoglu, Noam Shazeer, Oriol Vinyals
40 min
1 h 18Why GM will give you Gemini — but not CarPlay
WithMary Barra, Sterling Anderson
2 h 13
1 hGoogle Researcher Shows Life "Emerges From Code" - Blaise Agüera y Arcas
WithBlaise Agüera y Arcas
44 minQuestions people ask
- When did the Veo project begin?
- The Veo project's origins trace back to early research in 2018, with its beginnings in Google Brain.
- What are some challenges in video model development?
- Challenges include technical hurdles in evaluation and scaling, maintaining coherence in long-duration videos, and incorporating user feedback.
Page generated Oct 19, 2026. Summaries are Subvolts' own; the audio, artwork and notes belong to Google AI.