Rime voice models now available on Together AI

A voice agent can be correct and still feel broken. Customers judge it like a phone call: if it hesitates, sounds synthetic, or mispronounces a key term, trust collapses before they can evaluate reasoning. In production, that experience comes down to a real-time loop: STT (speech-to-text) models transcribe speech, the LLM decides what to say, and TTS (text-to-speech) speaks the response. At scale, teams stitch that loop across multiple vendors, so latency, reliability, observability, and ultimately what the customer hears become difficult to manage end-to-end.

Starting today on Together AI, the AI Native Cloud, we're adding Rime Arcana v2 and Mist v2 to the Together Model Library, bringing proprietary TTS models into the same API, authentication, and observability surface you already use for LLM and speech workloads. Arcana v2 delivers expressive, conversational voices trained on real customer service interactions, with 40+ voices across multiple languages and regional dialects for quality-critical scenarios. Mist v2 brings deterministic pronunciation control to high-volume production environments, reaching about 225ms time-to-first-audio on Together AI dedicated endpoints—you define how a term sounds once via API and it renders consistently across all voices, flows, and channels. Both run as dedicated endpoints on a single cloud alongside your LLM and STT workloads, so your end-to-end voice stack operates on one production platform — instead of being split across multiple providers.

Arcana v2: Expressivity for enterprise conversations

Arcana v2 is deployed today from high-growth startups to Fortune 500s as part of their production infrastructure. Across these environments, customers report measurable gains including 15% lift in sales at a national restaurant chain, a 75% reduction in call abandonment at a telecom provider, and a 10% increase in call success rates.

Trained on the largest proprietary dataset of full-duplex conversational speech data

Arcana v2 is trained on real conversations with everyday people — not audiobooks, podcasts, or voiceover announcers. The model learns natural breathing, fillers, backchannel cues, and conversational pacing from production conversations. Callers recognize these patterns and stay in the automated flow longer, improving completion and containment rates.

40+ voices and regional dialects

Arcana v2 ships with more than 40 voices across English, Spanish, French, and German. English includes 18 voices spanning U.K., Australian, and Southern US accents. Spanish includes four primary and three bilingual voices. Everyday words match local usage automatically. For example, "schedule" is pronounced "SHED-ule" in U.K. English and "SKED-ule" in U.S. English.

Mist v2: Deterministic pronunciation at production scale

Mist v2 is designed for high-volume production environments where pronunciation accuracy must be guaranteed across millions of calls. It already powers tens of millions of production calls each month for customer service and IVR systems where downtime or quality regression has direct revenue and compliance impact..

Deterministic pronunciation control

Most TTS models guess pronunciation on each generation. Mist v2 is deterministic. You define how a word should sound once through the API, and that pronunciation holds across more than 40 voices, flows, and channels. No retraining and no per vendor hacks. When your agent mispronounces a product name, drug, or acronym, you correct it once and the fix applies everywhere. Deterministic pronunciation configuration for Mist v2 is available today through our Sales team for production deployments; contact Sales to enable it for your environment.

English and Spanish with advanced pronunciation control

Mist v2 supports English and Spanish with deterministic pronunciation control. You specify how brand names, medication names, or technical terms should sound through the API, and Mist renders them consistently at conversational latency. If you need deterministic pronunciation at scale in Mist v2, contact Sales to enable it for your environment.

Proven at scale

Mist v2 serves tens of millions of calls monthly in production customer service and IVR environments. These are full-scale deployments where downtime or quality regression has direct revenue and compliance impact, not limited pilots.

Production-grade latency for conversational agents

Mist v2 reaches about 225ms p50 time-to-first-audio on Together AI dedicated endpoints. Voice agents need total end-to-end latency under 700ms to feel conversational, which means TTS must be fast enough to leave headroom for STT and LLM processing. When you co-locate Mist v2 with LLM and STT on Together AI, the entire pipeline from speech recognition through reasoning to synthesis stays within that budget, directly improving completion rates and user satisfaction.

Conversational realism

Like Arcana v2, Mist v2 is trained on real customer service calls. It preserves natural filler words, backchanneling, breathing patterns, and pacing while maintaining production-grade throughput. This makes it suitable for high-volume scenarios where both realism and responsiveness are required

Use cases

Global contact centers

Global teams can mix Arcana v2 and Mist v2 inside the same environment. Arcana v2 handles quality critical interactions like sales and complex support. Mist v2 handles high volume flows including basic inquiries and IVR routing. You can swap models with a configuration change, and keep configurations and observability unified through Together AI.

Real-time customer service

High-volume support flows need TTS latency under 250ms to feel conversational, with total end-to-end pipeline (STT → LLM → TTS) under 700ms. Mist v2 meets both thresholds when co-located with LLM and STT on Together AI, removing multi-vendor network overhead and keeping the pipeline inside a single environment.

Healthcare voice agents

Medication names like "lisinopril," "atorvastatin," and "metformin" must be pronounced correctly every time. Mist v2 uses deterministic pronunciation, so you define these terms once and they render correctly across 40+ voices. Running on Together AI HIPAA-compliant infrastructure means a single compliance review can cover the full voice stack.

Voice banking

Account numbers, routing numbers, and product names need to be read clearly and consistently across millions of calls. Rime’s models are trained on customer service conversations and are built for these high-precision use cases. On Together AI, banks and financial institutions can deploy Rime’s TTS models on SOC 2 Type II and PCI compliant infrastructure that meets their regulatory  requirements.

Production infrastructure on Together AI

Both Rime models run on Together AI Dedicated Endpoints on isolated GPU capacity alongside LLM and STT workloads. Together AI offers the broadest TTS catalog on a single platform, from open-source models like Orpheus and Kokoro to elite proprietary models like Rime, all with unified tooling.

The platform is built for production AI, with:

Get started

→ Try both models now

→ Read TTS Documentation

→ Contact Sales for deterministic pronunciation control, dedicated deployment, and volume pricing