Customer care teams can now deploy voice agents that sound natural, answer in real time and run where their data needs to stay. Gradium's voice models run natively inside Yampa, the agentic platform enterprises use to build, deploy and supervise customer care agents. The joint offer serves banks, insurers, mutual health insurers, telcos, energy, transport and public sector organizations, where control over data and suppliers is a regulatory requirement or a strategic decision.
What the partnership brings to customer care teams #
- Gradium models, orchestrated by the Yampa Voice Engine. Gradium Text-to-Speech and Speech-to-Text run natively inside the Yampa Voice Engine, Yampa's proprietary technology that orchestrates every call end to end, audio chunk by audio chunk. It handles turn-taking, silences, interruptions and background noise across every model in the loop. Gradium gives the agent its voice, Yampa gives it the timing of a real conversation. All in one platform, with no separate voice stack to assemble.
- Deployment on your terms. Each organization decides where its voice agents run: SaaS on European infrastructure, or on-premise in its own environment. Data and customer conversations stay within a perimeter the organization controls. Gradium is ISO 27001 and SOC 2 Type II certified. Yampa is SOC 2, hosted in the EU and GDPR compliant.
- One feedback loop between both teams. Gradium and Yampa technical teams work directly together, so what they learn on customer calls turns into model improvements within weeks.
What your customer care teams gain with Yampa #
Gradium makes the agent sound right, when Yampa makes it resolve the request behind the call.
- One command center for every agent. Yampa lets teams design, deploy, operate and govern every agent from one platform, with a consistent experience across voice, chat, email, SMS and messaging, full performance visibility and precise cost control.
- From blueprint to production in weeks. Teams build agents with a visual workflow builder, reusable components and blueprints, and test them with versioning before going live. Native integrations with CRMs, helpdesks and telephony let agents trigger real business actions during the call.
- Supervised on every call. Yampa tests agents before each update, monitors their decisions in real time and flags edge cases for correction. Teams track KPIs and audits live, run agents autonomously or with Yampa's Deployment Strategists, and scale to more than 100 agents.
- Results from day one. Yampa clients resolve 65% of after-hours calls autonomously, cut escalation rates by 40% in two months, and handle thousands of level-1 tickets each month without human intervention.
What Yampa voice agents gain with Gradium voices #
Each failed call starts in the voice layer: a slow answer, a misread order number, a caller who asks for a human operator. Gradium is built to remove those failures in production.
- Ultra-low latency. Gradium starts talking in under 50 milliseconds, with 0 ms of leading silence. The agent answers instantly, and latency and audio quality stay stable as call volumes scale.
- Reliable pronunciation of structured data. Gradium reads phone numbers, email addresses, dates, times, amounts in euros and alphanumeric confirmation codes correctly. Language-aware rewrite rules and custom pronunciation dictionaries give teams control over brand names, product references and specialist terms.
- Natural French with a native accent. Gradium delivers the same performance in French as in English or any other language, including regional accents such as Québécois French, so callers hear an agent that speaks their language the way they do.
- Robust voices designed for agents. Gradium voices are built for live conversation: they stay natural, clear and consistent across thousands of calls, whatever the request, the caller or the time of day.
A voice that belongs to your brand #
A custom voice becomes part of a company's brand identity, alongside its logo and visual design. The project brings in marketing, communication and brand teams next to customer service, and the choice of voice moves to the executive committee.
Gradium gives brand teams two ways to create that voice. With Voice Designer, a team describes the voice it wants in a text prompt (age, tone, accent, energy) and gets it back in a few seconds. Teams iterate on the prompt until the voice matches the brand, then deploy it in their Yampa agents. With voice cloning, a company reproduces an existing voice, such as a brand ambassador's or the voice already used in its ads. The same voice then carries across phone calls, campaigns and every language the agent speaks.
For organizations that want to go further, Gradium and Yampa run a curated process together to optimize the voice for a specific use case. The teams test the voice on real scenarios, such as claims handling or payment reminders, and fine-tune tone, pace and pronunciation until it performs on the calls that matter most.
What both teams say #
"At scale, a voice agent is only as good as the hardest word it has to say. License plates, policy numbers, URLs, medical terms: Gradium's models, pronunciation dictionaries and rewriting rules get them right instantly, in each industry's vocabulary and in a voice that's truly the brand's own. Their team brings a researcher's rigor and an artist's ear to voice, and our clients hear the difference on every call."
Marin Huet, CEO, Yampa
"Yampa puts our models to the test every day on real customer calls, for some of the most demanding organizations in customer care. We work hand in hand with their team to turn that feedback into model improvements within weeks, so Yampa's clients always run the most robust and reliable voice agents in production."
Neil Zeghidour, CEO, Gradium
About Yampa
Yampa is a Paris-based company founded in 2024 by Marin Huet, Baptiste Saintot and Patrice Mazoyer, and backed by Partech. Its agentic platform lets enterprises design, deploy and supervise autonomous agents that resolve customer requests across phone, email and chat, and escalate complex cases to human teams. Yampa serves more than 50 enterprise and tech-native clients and handles 400,000 interactions per month. yampa.ai
About Gradium
Gradium is a Paris-based company founded in 2025 by Neil Zeghidour, Laurent Mazaré, Olivier Teboul and Alexandre Défossez, spun out of the Kyutai research lab and backed by NVIDIA. It builds the voice layer for real-time applications, with models optimized for voice agents: Text-to-Speech (TTS), Speech-to-Text (STT), voice design, voice cloning and live translation. gradium.ai
Get started #
Organizations evaluating voice agents for customer care can contact Yampa at yampa.ai or talk to us.
Related posts #
Sep 9, 2026AnnouncementDiscover Gradium now available for free on LiveKit Inference
Gradium Text-to-Speech is now available on LiveKit Inference, free until Oct. 9. Gradium TTS is designed for conversational use cases, with voice cloning, low latency (below 250ms TTFA), and robust pronunciation on difficult cases.
Sep 7, 2026ProductDiscover Launching Voice Design: prompt the voice your agent needs
Developers building voice agents ask us for voices matched to the use case in front of them: a Québécoise receptionist for a Montréal dealership, a Paulista support agent for a São Paulo fintech, a narrator in his sixties with the authority of a lecture hall. Briefs outnumber any catalog. Voice Design starts from a prompt: write one or two sentences, get complete new voices back in seconds, keep the one you want.
Sep 2, 2026ProductDiscover Build voice agents with EU and US residency and Zero Data Retention
Two questions come up in almost every enterprise evaluation: where exactly does the inference run, and what do you keep? Data residency pins your Text-to-Speech, Speech-to-Text and Speech-to-Speech sessions to our EU servers, or to our US servers, enforced at the server and reported on every response, so you can assert on it instead of trusting a hostname. Zero Data Retention decides whether the text and audio processed by our servers survive the call at all.