Voice AI Startup Rime Raises $24M to Fix Enterprise Calls

SAN FRANCISCO: Rime raises $24 million in Series A funding to make AI voices stop butchering brand names on customer calls.

M13 leads the round. Unusual Ventures and existing backers join. The raise follows a $5.5 million seed from May 2025. Enterprises use Rime’s models to field phone calls at scale.

The company’s training approach is genuinely odd for this industry. Instead of scraping web audio like everyone else, Rime built an actual recording studio in San Francisco. Voice actors record conversational sessions there, and that studio data trains the models.

Also read: Lema AI Raises $24M to Fix Enterprise Supply Chain Security

Also read: Nvidia Backs French Voice AI Startup Gradium at $100M Seed

The payoff is pronunciation. A phoneme-based system handles accents and regional variation, and the models say product names and industry jargon correctly. That sounds trivial until an automated agent mangles your company’s flagship product on ten thousand calls a day. Mispronunciation remains one of the most common failures in enterprise voice deployments.

Also read: WINN.AI Raises $18mn Series A to Guide Sales Reps Live on Calls

Founder Lily Clifford is refreshingly blunt about the category’s limits. Most current deployments, she argues, are just better-sounding versions of the old IVR phone tree. Nicer voices, same rigid experience. Fixing that gap is the actual product roadmap.

Money is pouring into voice right now. OpenAI’s new GPT-Live pushes simultaneous listening and speaking to 150 million weekly voice users. Restaurants, clinics, and service businesses want reliable booking and triage calls, and generic models keep stumbling on their vocabulary.

Rime’s bet is that domain-specific training data wins where scale alone hasn’t. The proceeds fund more studio recording, enterprise sales, contact-center integrations, and new language coverage. Whether a startup can out-collect the giants on data is the question the next 18 months will answer.