Editorial illustration for Google hires Hume AI staff to add voice, emotion to DeepMind models
Gemini 3: Google's Most Intelligent Search AI Yet
Google hires Hume AI staff to add voice, emotion to DeepMind models
Google's DeepMind just hired Alan Cowen, the founder of Hume AI. That's the hard fact. It signals a sharp pivot: the lab is now embedding voice and emotion-reading tech directly into its core models.
The startup itself burned millions training systems to parse subtle emotional cues in recorded speech. And the new Hume CEO, Andrew Ettinger, states it plainly—voice is headed for primetime as the way people talk to AI.
Cowen and the other Hume AI recruits will help Google DeepMind integrate voice and emotional intelligence into its latest models, according to sources who spoke on the condition of anonymity as they aren't authorized to speak publicly about the deal. Hume AI has invested millions in developing models and tools to hone realistic voice interfaces and to detect emotions in the voices of users. The company trains its models by having experts annotate emotional cues in real conversations.
At Google, Cowen and his colleagues will help the tech giant integrate voice and emotion technology into its frontier models, sources say. "Voice is going to become a primary interface for AI, that is absolutely where it's headed," says Andrew Ettinger, an experienced investor and executive who is taking over as the CEO of Hume AI.
This is a strategic shift for Google. The focus moves beyond raw data processing toward making AI sound, and feel, human. Their specific goal?
Building models that detect subtle vocal tones—frustration, fatigue, everything. If Ettinger’s prediction holds and voice becomes the main interface, this underlying emotional tech will define how millions finally engage with Google's products.
Related Reading
Browse all Google news →
Business & Startups Nordic pilot adds Gemini for Education, NotebookLM to boost AI literacy
Business & Startups Kling launches Video O1, all-in-one model with MVL bridge using transformer
Business & Startups DeepSeek Seeks More Capital Weeks After USD 7B Funding Round
Related Topic Google tests visual 'magazine-style' UI for Gemini 3 Pro users
LLMs & Generative AI
Related Topic Google's FACTS benchmark shows 70% factuality ceiling across four tests
Research & Benchmarks
From Archives Meta AI lab to debut ‘Avocado’ text model and ‘Mango’ vision model in Q1
From Archives Telangana Launches Aikam AI Innovation Hub at Davos, Calls for Global Partners
From Archives India aims for USD 150 bn AI spend by 2026; Google, Microsoft, Amazon commit USD 70 bn
From Archives Free on-demand full-length SAT practice tests now live in Gemini
LLMs & Generative AI
Common Questions Answered
What is OCTAVE and what makes it unique among speech-language models?
OCTAVE (Omni-Capable Text and Voice Engine) is a next-generation speech-language model that can generate not just voices, but entire personalities from brief prompts or recordings. Unlike traditional text-to-speech systems, OCTAVE can create multiple interacting AI personalities with distinct characteristics like gender, age, accent, and emotional intonation, while maintaining the capabilities of a frontier large language model.
How detailed can OCTAVE's voice and personality generation be?
OCTAVE can generate extremely nuanced voices and personalities with remarkable specificity, from a 'gravelly male voice as if gargling hot asphalt' to a 'New Zealand female wellness coach with a soothing, deliberately slow therapeutic voice'. The model can emulate precise vocal characteristics including vocational speaking styles, emotional tones, and even specific accent variations with professional-level detail.
What are the key capabilities of OCTAVE for developers and creators?
OCTAVE is designed to power AI systems that can communicate richly with humans while following detailed instructions and using tools. The model is well-suited for applications like creating multi-character audiobooks, generating podcast dialogues, producing video voiceovers, and developing conversational agents with highly customizable vocal and personality characteristics.
Further Reading
- Papers with Code - Latest NLP Research — Papers with Code
- Hugging Face Daily Papers — Hugging Face
- ArXiv CS.CL (Computation and Language) — ArXiv
