Rime just closed a $24 million Series A. It's betting enterprises are finally ready to let an AI voice run a live customer phone call, not just staff a chatbot.
If you've ordered a pizza from Domino's or grabbed wings from Wingstop by phone this year, there's a real chance the voice that took your order was synthetic.
That voice likely came from Rime, a San Francisco startup that builds text-to-speech models tuned for real phone calls rather than scripted demos. Rime raised $24 million in Series A funding led by M13, with Twilio Ventures and Corazon Capital joining alongside existing backer Unusual Ventures, according to funding data compiled by PitchBook and Crunchbase. The company's speech models now carry more than 100 million phone conversations a month for more than 20 enterprise customers. That's not small.
It's a fast follow-up. Only fourteen months earlier, in May 2025, Rime raised a $5.5 million seed round that Axios reported was led by Unusual Ventures, alongside Founders You Should Know and Cadenza Capital. If you're trying to understand why investors came back so quickly, don't start with the usual voice-AI demo reel. Start with the phone line.
Rime was founded in 2022 by Lily Clifford, who left a computational linguistics PhD program at Stanford to start the company. She was joined by Brooke Larson, a language engineer who came from Amazon's Alexa team, and Ares Geovanos, who had worked at UC San Francisco building brain-computer interfaces for patients who had lost the ability to speak. That background shows up in the product. A call center voice model has to survive bad phone lines, cross talk, and background noise. No lab demo has to do that. Geovanos and his co-founders have spent years thinking about what makes a voice recognizably human under pressure, not just intelligible in a quiet room.
The Phone Call Is the Product
Business Insider reported in May 2025 that Rime's text-to-speech technology was being used by Domino's and Wingstop for phone ordering, with Domino's using the feature in about 80% of its North American phone orders, according to Clifford. The detail that sticks isn't the percentage alone. It's the reason customers stayed on the line. Clifford told Business Insider the voice should sound like someone who could work at Domino's, not a 20th-century radio announcer.
That is the right instinct. People don't call a pizza shop looking for a showcase of emotional range. They want the order right, the menu item pronounced correctly, and the call over quickly. Business Insider also noted that Rime recorded people talking naturally with friends and family in a San Francisco studio rather than relying only on voice actors reading polished scripts. The result is less showroom audio and more working audio, which is exactly what a restaurant chain needs when the caller is in a car, a kitchen, or a noisy room.
Latency Is the Real Gatekeeper
Two years ago, handing a live customer call to a model felt reckless to most brands. A wrong answer in a chat window can be edited before anyone notices it. A wrong answer on a phone call, spoken in a voice a customer already trusts, is a different kind of risk entirely.
Speed changed. The important shift in voice AI isn't only that models sound smoother. It's that the delay between a customer's words and the machine's reply has fallen enough for the exchange to feel less broken. An arXiv technical tutorial published in March 2026 described the current industry pattern as a streaming pipeline: speech-to-text, then a language model, then text-to-speech, with each stage passing output along instead of waiting for the whole answer to finish. That sounds technical, but you feel it immediately. Pause too long and the caller knows.
Rime's narrow bet is that phone calls deserve their own stack, not a generic synthetic voice bolted onto whatever agent a company already uses. That gives it a sharper story than most startups in the space. It also gives it less room to hide. If the voice misses a name, mangles a limited-time menu item, or sounds too cheerful for a drive-thru at rush hour, the customer hears the failure in real time.
ElevenLabs Is the Shadow Over the Room
Rime is walking into a market with an obvious giant already in it. The Wall Street Journal reported in February 2026 that ElevenLabs raised $500 million at an $11 billion valuation and generated $330 million in recurring revenue in 2025, with plans to double that in 2026. That's a hard market. ElevenLabs isn't just a voice clone company anymore. It has pushed into customer service, onboarding, marketing, dubbing and voice licensing, and it has the capital to chase enterprise phone calls if that becomes the prize.
Rime hasn't tried to out-market ElevenLabs. Its pitch is narrower: build for the call itself, not for every use case a voice model could serve, from audiobook narration to game characters, you name it. Frankly, that narrowness is the strongest part of the company. A restaurant chain routing thousands of calls a day through Rime cares less about how expressive its voice sounds reciting a poem and more about whether it gets an order right when a customer mumbles through a bad connection.
None of this guarantees Rime outlasts its better funded rivals. ElevenLabs and Deepgram both have the resources to build call center-specific products of their own, and Twilio, now one of Rime's investors through Twilio Ventures, already runs infrastructure that carries a huge share of business phone calls. That's either useful distribution or a warning sign, depending on how Twilio chooses to compete. For now, the $24 million buys Rime time to widen its enterprise base before larger companies turn their attention to the same phone lines Rime built for first.
Also read: Indian Coding Startup Emergent Becomes a Unicorn in Just Over a Year • Kevin Ryan's AlleyCorp Raises $335 Million and Still Won't Chase Mega-Rounds • Anthropic Gives Teachers Free Claude Access as AI Giants Fight for Classrooms
