Gradium
Gradium builds the foundational AI models and infrastructure for voice: real-time speech-to-text, expressive text-to-speech, voice generation, translation and conversational intelligence. Its stated aim is to translate frontier audio research into production-ready systems for developers and enterprises building natural, low-latency voice experiences.
At a Glance
- Developers and startups building voice agents
- Enterprise customer-experience and contact-center teams
- Healthcare and accessibility providers
- Media and broadcasting
- +3 more
AI Tools by Gradium
(1)Phonon
On Device Text to Speech Models
Discussions
No discussions yet
Be the first to start a discussion about Gradium
Latest News
A Closer Look at Phonon: launch story, demos and Open TTS Leaderboard performance
Gradium and Yampa partner on voice AI agents for customer experience
Gradium releases Phonon, an open-source TTS model that runs on any device
Open-sourcing Phonon: suite of frontier on-device TTS models
Products & Services
Developer and enterprise platform for streaming real-time voice interactions, with WebSocket/API access to speech models.
Expressive streaming TTS with timestamp accuracy, custom pronunciation, multilingual support and scalable concurrency; the current product page reports under-50ms time to first audio and hard-case handling for phone numbers, emails, acronyms and codes.
Real-time transcription with code-switching, controllable latency, custom pronunciation and semantic voice-activity/turn detection for conversational agents.
Low-latency speech-to-text and speech-to-speech translation across English, French, German, Spanish and Portuguese, with control of output voice including cloning.
Market Position
Gradium positions itself as a foundational, real-time voice AI infrastructure provider emphasizing very low latency, reliable pronunciation of difficult enterprise text, multilingual streaming models and production scale. Its published TTS comparison names Cartesia, ElevenLabs and Inworld among real-time competitors; broader market competitors include Google/Gemini and AssemblyAI. Phonon differentiates through small open-weight models that run privately and offline on consumer/edge hardware.
Leadership
Founders
Neil Zeghidour
Co-founder and CEO; previously led audio/voice research at Google Brain, Google DeepMind and Meta FAIR; co-founded Kyutai.
Laurent Mazaré
Co-founder and Chief Coding Officer; former CTO of Kyutai, with prior experience at Google DeepMind and Jane Street.
Olivier Teboul
Co-founder and Chief Technology Officer; previously worked at Google Brain and Kyutai on audio models and infrastructure.
Alexandre Défossez
Co-founder and Chief Science Officer; previously worked at Meta/Meta FAIR and Kyutai on audio research, codecs and multimodal speech-text models.
Executive Team
Neil Zeghidour
Co-founder and Chief Executive Officer
Audio/voice researcher formerly associated with Google Brain, Google DeepMind and Meta FAIR; co-founder of Kyutai.
Olivier Teboul
Co-founder and Chief Technology Officer
Former Google Brain researcher and Kyutai founding researcher focused on audio technology and infrastructure.
Founding Story
Gradium was spun out of the nonprofit AI research lab Kyutai in September 2025 by Kyutai's founding researchers. The founders created it to bring more than a decade of frontier research in audio codecs, audio language models and real-time speech systems into production-ready infrastructure and products for developers and enterprises.
Business Model
Revenue Model
Usage-based API/platform subscriptions: customers buy monthly credits for TTS, STT and translation, with pay-as-you-go extra credits and custom Enterprise plans; Gradium also offers a startup grant and free tier.
Pricing Tiers
45,000 credits/month.
225,000 credits/month; 1,000 custom voices.
900,000 credits/month; 1,000 custom voices.
9 million credits/month; 1,000 custom voices and 5 Pro Voice Clones.
45 million credits/month; 1,000 custom voices and 20 Pro Voice Clones.
Unlimited credits and custom concurrency/voice terms.
Target Markets
- Developers and startups building voice agents
- Enterprise customer-experience and contact-center teams
- Healthcare and accessibility providers
- Media and broadcasting
- Automotive and connected-vehicle applications
- Consumer applications and gaming
- Voice agents and conversational AI
- Customer experience and contact centers
- Healthcare and accessibility
- Media, radio and personalized audio
- Consumer apps and offline products
- NPCs and video games
- Renault / RMC BFM Drive
- Wonderful
- Acolad
- Yampa