Two models designed to keep customers on the line.
Two TTS APIs that have everything you need to power voice agents that perform at enterprise scale.
Coda
Flagship · Default
Choose Coda when quality is the point. Coda holds the highest scores in our human evals when listeners are asked which voice they'd prefer to speak with.
<100ms
Model latency
184
voices
Mist v3
fastest
Choose Mist when milliseconds matter. At ~37ms to first audio, plus control over pauses and pronunciation, Mist is built for scaling conversations for high stakes, and high volume.
~37ms
Model latency
78
voices
Coda
flagship
High-stakes customer experience calls
Sub-100ms model latency
8 languages · 184 voices
Word timestamps · spell() · speed
Cloud · On-prem · Co-located endpoint
best for
latency
languages/ voices
controls
deployment
Mist v3
fastest
Hard real-time budgets; regulated flows
~37ms p50 time-to-first-byte
4 languages · 78 voices
Custom pauses · spell() · speed
Cloud · On-prem · Co-located endpoint
Real scenarios, real voices.
Hear how Rime voices come to life for different use cases
Coda voices in real-world scenarios
professional
Cupola
calm
Eliphas
happy
Astra
casual
Vespera
professional
Cupola
calm
Eliphas
happy
Astra
casual
Vespera
professional
Cupola
calm
Eliphas
happy
Astra
casual
Vespera
professional
Cupola
calm
Eliphas
happy
Astra
casual
Vespera
What real callers think about Rime voices
Call completion
Rime voices keep customers on the line longer.
Rime
12.1%
ElevenLabs
10.3%
Google
9.7%
Callers still on the line after hearing they were talking to an AI. Miravoice, 99,400 calls, 2026. Read the study
Quality
Rime beats ElevenLabs when it comes to a voice that customers want to engage with.
Rime
67%
ElevenLabs
21%
No preference
13%
Which sounds more natural, the same lines on both. 350 blind ratings by third-party listeners (Podonos), Coda against ElevenLabs Flash 2.5.
From the teams that put us on the line
"Mist v3 on a co-located endpoint has been a 3x latency improvement for us. We're consistently seeing sub-100ms TTFB in production — that's the difference between a conversation that feels human and one that doesn't."
Ali Mansoor
Founding Engineer, Trillet AI
"The latency stays low even under heavy loads. It's a true enterprise production-grade system."
Pratik Mundra
Co-Founder, SigmaMind AI
"It was clear that the voice made a real impact on the phone. When guests feel comfortable right away, everything else goes more smoothly."
Rahul Aggarwal
Founder & COO, ConverseNow
"In healthcare, the voice is the first point of trust. Patients need to feel heard, not processed — and that starts with how the AI sounds and how fast it responds."
Jack Ryan
Co-Founder & Chief Product Officer, Attune
.svg){kind=logo}
"Mist v3 is incredibly fast and deterministic pronunciation is a game-changer. When you're powering live conversations at scale, you can't have a model guessing brand names or proper nouns."
Tom Shapland
Product Manager, LiveKit
Enterprise features
What it takes to run a voice in production, not a demo.
Runs where your data lives
Cloud, VPC, on-prem, or a co-located endpoint.
BAA and SOC 2
HIPAA-ready with a BAA; SOC 2 reports on request.
600+ unique voices
We have the largest dataset of conversational voices to choose from.
Pronunciation you control
spell(), custom rules, and word timestamps, so names and IDs come out right.
SLAs and a named team
Uptime commitments and dedicated support.
Observability
Per-call latency and QA, so a regression is caught rather than heard about.
Questions, answered
How do i choose between Coda and Mist?
Start with Mist if latency is the constraint and Coda if expressiveness is. Both share the API, so testing the other one is a one-line change.
Can I bring my own voice?
Yes. Custom voice clones are trained from your own recorded talent, and enterprise plans include unlimited clones.
What does deployment look like?
Cloud by default. Enterprise plans can run in your VPC or fully on-prem, which is how teams under HIPAA or similar constraints deploy Rime.