Rime builds voice models for enterprise phone systems, and the San Francisco startup has raised a $24 million Series A to push deeper into AI-powered calls. The pitch is simple: most enterprise voice bots still sound off, pause too long, or butcher brand names and industry jargon. That's why big companies keep clinging to old IVR setups. Founded in 2022 by Lily Clifford, Brooke Larson, and Ares Geovanos, Rime is betting that better conversational speech data — collected in its own studio, not scraped from the web — can make phone automation feel less synthetic and more useful.
What is Rime voice AI and how does it work?
Rime voice AI is a text-to-speech platform for real-time voice agents and IVR systems. A customer picks a model, chooses a voice, sends text through Rime’s API or dashboard, and gets back streaming speech designed for live phone conversations. The default choice for most new deployments is Coda, while Mist is aimed at teams that care most about speed or deterministic pronunciation behavior.
The workflow is pretty practical. Teams can connect over standard HTTPS or WebSockets. They can route traffic to the nearest regional endpoint for lower latency, and self-host or run on-prem when compliance rules demand tighter control. Rime also exposes word-level timestamps. Those matter more than they sound. They help with interruption handling, barge-in, and the kind of turn-taking that keeps a call from feeling awkward.
What removes manual work is the tuning layer around speech itself. Rime offers 600+ voices and 50+ languages and dialects. It also includes controls for accent, pace, and tone, plus tools for spelling names, numbers, and codes out loud. Mist v2 is the model built for fine-grained pronunciation control, which lines up with the company’s broader pitch: customers shouldn’t have to retrain a model every time a brand name, place name, medical term, or fintech acronym comes up on a call.
That’s where Rime is trying to stand apart. Its models are trained on full-duplex conversational speech between real people, with interruptions, laughter, hesitation, and all the messy bits that usually get cleaned out of training data. Coda is built for sub-100ms model latency when self-hosted or on-prem, and cloud deployments add roughly 25–50ms of network round-trip time across most of the continental U.S. Small gap. But it's the difference between a voice that feels responsive and one that sounds like it's waiting for permission to talk.
Who founded Rime and why build voice AI for enterprises?
The founding story
Rime was founded in 2022 by Lily Clifford, Brooke Larson, and Ares Geovanos. Clifford had been a Stanford PhD student in computational linguistics. Larson came in as a PhD linguist with Amazon Alexa experience. Geovanos brought the engineering and product side. Together, they set out to build speech that sounded less polished in the artificial sense and more human in the useful sense.
The company’s earliest strategic choice still explains a lot about it. Instead of scraping random audio from the internet, Rime built a recording studio in San Francisco and started collecting its own conversational dataset. That dataset became the base for models trained on spontaneous, overlapping, real-world speech — not audiobook cadence, not voiceover cadence, not demo-day cadence.
Why this team fits the category
Clifford’s background is unusually relevant for this market. She has talked publicly about studying computational linguistics and sociophonetics at Stanford before leaving to build Rime, which helps explain why the company talks so much about rhythm, stress, dialect, and trust rather than just raw speech generation. Brooke Larson’s Alexa work adds the hard-earned lesson that voice systems fail fast when they don’t understand how people actually speak.
Geovanos adds another piece that isn’t trivial. Before Rime, he was working at UC San Francisco on brain-computer interfaces for people who had lost the ability to speak. Different domain. But it suggests the founding team wasn’t coming at speech as a toy problem. They were already thinking about communication as infrastructure.
Early traction and the Series A
Rime is already live through its dashboard and API, and it powers tens of millions of conversations each month. Its customer base spans food service, healthcare, airlines, and fintech, and the startup has cited enterprise names including Mayo Clinic, Dialpad, Upstart, and Asurion. On the product side, it has also been pushing volume: the homepage says Rime powers more than 1.5 million minutes of conversation.
M13 led the Series A, with participation from Twilio Ventures, Corazon Capital, Unusual Ventures, and other existing backers. It follows a $5.5 million seed round announced on May 29, 2025. Morgan Blumberg from M13 is joining the board, and Rime has also added Rafael Valle as chief scientist after his work at Meta Superintelligence Labs and Nvidia’s applied deep learning audio research team.
How Rime plans to use the funding
Rime will use the new money to expand beyond its current team of 35, with hiring aimed at model development, engineering, and partnerships. It also marks a technical shift. The company started with separate speech-to-text, text-to-speech, and LLM components. Now it’s shifting toward speech-to-speech models to cut latency, improve turn-taking, handle background noise better, and reduce how much orchestration glue it has to manage.
Competition is crowded. On one side, you’ve got voice model developers like ElevenLabs and Deepgram. On another, infrastructure players like Vapi, Retell, and LiveKit. Then there are full-stack customer support companies like Decagon and Sierra. And hanging over all of them is the oldest incumbent in the room: IVR.
Clifford isn’t pretending the problem is solved. She said, “The voice technology is still not there to automate the vast majority of enterprise phone calls. LLMs have made it a lot easier to build voice applications that work, but they haven’t changed how it feels to interact. Talking with a voice AI agent is not the most compelling experience for the end user. It’s kinda like a new IVR, but with a better voice.”
That honesty is part of the positioning. Rime isn’t selling a fantasy about totally solved voice automation. It’s selling lower latency and stronger pronunciation control. It also has proprietary conversational data and deployment options that fit regulated environments. Blumberg put it this way: “Companies like ElevenLabs have moved into being an orchestration and the application layer, going head to head with the Sierras and Decagons of the world. I think there’s just so much more to be done technically, and Rime’s approach of pushing forward on the best model with low latency and high reliability in a regulated environment stands out.”
Why does this $24M round matter for Rime voice AI?
This isn’t just growth capital. It’s model-building capital.
Rime is using the round to hire deeper into research and engineering at the exact moment it’s trying to move from a stitched-together pipeline toward more capable speech-to-speech systems. If that works, the payoff isn’t just nicer audio. It’s less latency and smoother interruption handling. Better performance in noisy environments too. It also means less architectural overhead from coordinating too many separate models.
It also matters for customers in regulated or high-trust settings. Healthcare, fintech, and travel don’t need a fun demo voice. They need something fast, accurate, and predictable enough to sit in front of patients, cardholders, or travelers without immediately eroding trust. Rime’s on-prem and self-hosted options make that thesis more believable than a cloud-only pitch would.
How big is the enterprise voice AI market?
Pretty big already, and still climbing fast. Grand View Research estimates the conversational AI market at $17.6 billion in 2026 and projects it to reach $78.9 billion by 2033. In the narrower call center AI segment, the firm expects the market to hit $7.08 billion by 2030, growing at a 23.8% CAGR from 2025 to 2030.
The more interesting shift is structural. Enterprises are no longer buying “voice AI” as one monolithic thing. They’re picking across layers — model vendors, orchestration stacks, and application companies — while still benchmarking all of it against legacy IVR. That’s why Rime’s bet on speech quality, pronunciation, and latency matters now. Building a voice app is easier than it was 2 years ago. Building one that people don’t hate on a real customer call is still hard.
Can Rime voice AI beat IVR in enterprise calls?
Rime has a credible shot because it’s not trying to win with a generic “AI agent” story. It’s going after the parts of voice automation that still break in production — speed, pronunciation, turn-taking, and trust. If the company can turn its studio-trained data and speech-to-speech push into measurably better live calls, Rime voice AI could become an important layer in enterprise phone systems. If not, it risks becoming exactly what Clifford warned about: just a better-sounding IVR.
Read how Promom raised ₹30 crore from Fireside Ventures to expand its maternal and baby care products, grow nationwide distribution, and build feeding solutions designed for Indian mothers.
FAQ
- What funding did Rime just raise? Rime raised a $24 million Series A announced on July 15, 2026. M13 led the round, and Morgan Blumberg joined the board as part of the deal. The new financing came a little over a year after Rime’s $5.5 million seed round in May 2025.
- How does Rime’s voice AI product work? Rime gives companies a dashboard and API for real-time text-to-speech in phone calls and voice agents. Teams choose a model like Coda or Mist, configure voice and pronunciation behavior, then deploy through standard API connections or on-prem if they need tighter compliance controls. It also supports streaming output and word-level timestamps, which help live systems react naturally during interruptions.
- Who are the founders of Rime? Rime was founded in 2022 by Lily Clifford, Brooke Larson, and Ares Geovanos. Clifford came out of Stanford’s computational linguistics world, Larson previously worked on Amazon Alexa, and Geovanos brought engineering and product experience that included brain-computer interface work at UC San Francisco. It’s a founding team with real speech and language depth, not just general AI credentials.
- Is Rime a call center AI company or a voice infrastructure startup? It sits closer to the voice model and infrastructure layer than to a full outsourced contact-center platform. Rime sells the speech layer — voices, latency, pronunciation, deployment, and APIs — that other enterprise products can build on top of. That’s why its competitive set stretches from model makers to orchestration vendors, even though its core bet is still the underlying voice technology itself.
