Last Updated: 15 July 2026
Rime, a San Francisco startup that applies linguistics research to AI voice technology, has raised $24 million in Series A funding. The company builds voice models for enterprises in healthcare, finance, and customer service, training them on recorded human speech rather than web-scraped audio. Lily Clifford, who left a Stanford linguistics PhD to co-found the company in 2022, leads Rime as CEO.
The round comes at a time when businesses are rapidly adopting AI to handle phone calls. The global AI voice agents market was valued at $2.54 billion in 2025 and is projected to grow at a compound annual rate of 39% through 2033, according to Grand View Research. Healthcare is the fastest-growing segment, with a projected 42% annual growth rate over the same period. As companies replace older phone systems with AI voice agents, getting pronunciation right in regulated industries like medicine and banking has become a real technical problem.
M13, an early-stage venture capital firm based in Los Angeles, led the round. Twilio Ventures, Corazon Capital, and existing investor Unusual Ventures also participated. Morgan Blumberg, a partner at M13, will join Rime's board of directors.
Most voice AI today works well enough for simple tasks. It still struggles, though, in situations where small mistakes carry real consequences. Think of a healthcare call explaining medication dosages, or a financial services call walking someone through loan terms. A mispronounced drug name or an awkward pause in the wrong place can make a caller lose trust fast. Rime's argument is that sounding generically "natural" is not sufficient. The voice also needs to fit the specific context, pace, and tone a conversation requires.
An independent study by Miravoice offers some evidence for this. The study tested 12 voices from three vendors across 100,000 phone calls. Rime's voices produced significantly lower rates of callers hanging up during the introduction compared to other vendors in the study. Rime also showed the fastest median time to call completion among all providers tested, suggesting that cadence and pacing do affect whether people stay on the line.
"In financial services, the stakes for voice AI are exceptionally high and accuracy is a requirement for customer trust. Rime’s ability to handle complex financial use cases with precision allows us to build experiences that truly serve our customer’s needs."
Shiv Indap, Upstart
The new capital will go primarily toward two things: growing Rime's proprietary conversational speech dataset and hiring in engineering and research. The team currently numbers about 35 people.
Alongside the funding announcement, Rime named Rafael Valle as Chief Science Officer. Valle previously led audio research at Meta's Super Intelligence Lab and worked on applied deep learning audio research at Nvidia. His appointment points to where Rime wants to go next. The company is pushing into speech-to-speech models, meaning systems where the AI both listens and responds in real time rather than converting speech to text and back. Clifford has called the longer-term goal "conversational general intelligence" for voice, where the AI can adjust emotion, pacing, and emphasis on its own without explicit instructions.
"The design patterns that will define the speech-to-speech era of AI haven't been built yet. Rime is doing the foundational work by combining frontier AI with deep linguistic expertise to create the speech infrastructure the next generation of AI products will rely on."
Morgan Blumberg, Partner at M13
Three co-founders started Rime in 2022. Lily Clifford left her linguistics PhD at Stanford. Brooke Larson, a PhD linguist, came from Amazon's Alexa team. Ares Geovanis brought an engineering background from Stanford. All three share roots in academic speech science, which informs how Rime approaches voice AI.
Instead of scraping audio from the internet, Rime built a recording studio in San Francisco and recorded its own dataset of conversational speech. That dataset, which the company describes as one of the world's largest proprietary collections of expressive, multilingual conversational audio, includes the messy parts of real talk: interruptions, laughter, and filled pauses like "um" and "uh." It covers a wide range of American accents and dialects. The goal is to train models that can handle what real business calls actually sound like, from brand names and numerical sequences to medical terminology and financial jargon.
On the product side, Rime offers more than 600 voices in over 50 languages with response times under 100 milliseconds. It holds HIPAA compliance and SOC 2 Type II certification, and it can run on a client's own servers when data security rules require it. The company says it currently powers close to 100 million phone calls per month for clients including Mayo Clinic, Dialpad, Upstart, and Asurion.
"The things that matter most in conversation are the hardest to pin down. Response relevance is hard to specify; speech appropriateness is hard to verify. There's no unit test for sounding like you care. That's exactly why this is a design problem, not just an engineering one."
Lily Clifford, Co-founder and CEO of Rime
M13 led the Series A. The firm, established in 2016, focuses on seed and Series A investments in software companies and has made more than 200 direct investments, with multiple exits and several portfolio companies that reached unicorn status. Other participants in this round include Twilio Ventures, the investment arm of cloud communications company Twilio; Corazon Capital; and Unusual Ventures, which also led Rime's earlier seed round.
Before this Series A, Rime had raised $5.5 million in a seed round announced in May 2025. Unusual Ventures led that round, with participation from Founders You Should Know, Cadenza Capital, and a group of angel investors. Across its seed and Series A rounds, the company's disclosed funding totals approximately $29.5 million.
