Skip to main content
Start with Coda for new applications and Arcana migrations. Choose Mist v3 when the lowest time to first audio matters most, or when you need inline pronunciation control or custom pauses; prefer it over the older Mist v2. Arcana is deprecated and will be sunset on August 15, 2026; see the migration notes below.
Arcana will be sunset on August 15, 2026. Cloud traffic will route to Coda on that date. Migrate by changing modelId: arcana to modelId: coda; the API contract is unchanged. Sinhala remains available on Arcana. On-prem Arcana images remain available to pull and run, but will not receive updates or support (including vulnerability patches and text-normalization fixes) after August 15, 2026.

​
Choose a model

For benchmark methodology, latency, and throughput numbers, see Latency. All production models are available through the cloud API and on-premises, and all stream audio over HTTP and WebSockets. Rime currently operates coda, arcanav3, arcanav2, mistv3, and mistv2 model deployments.
Use the regional endpoint closest to your application to reduce network latency.

​
Feature matrix

The Mist column reflects Mist v3 and Mist v2. Both support pronunciation control and custom pauses.

​
Coda

Coda, released May 2026, is Rime’s flagship TTS model and the successor to Arcana. It pairs an LLM backbone with a dedicated speech inference engine trained on full-duplex conversational data.
  • Received the highest voice-quality scores in Rime’s human evaluations for naturalness, prosody, and artifact-free output
  • Sub-100ms model latency on the GPU engine when self-hosted or on-prem
    • Cloud API users typically add 25 to 50ms of network round-trip time from most of the continental US when routed to the closest regional endpoint
  • One voice lineup across English, Spanish, French, Portuguese, German, and Japanese
  • Word-level timestamps for text-audio alignment and interruption handling
  • Supports spell() for spelling sequences letter by letter or number by number
  • Available with modelId: coda

​
Arcana

Arcana is deprecated and will be sunset on August 15, 2026. On that date, cloud Arcana traffic will route to Coda. Migrate by changing modelId: arcana to modelId: coda; the rest of the API contract stays the same. Sinhala remains available on Arcana. On-prem Arcana images remain available to pull and run, but will not receive updates or support after August 15, 2026.
Arcana, released April 2025, is Rime’s previous flagship model.
  • 94 voices spanning age groups, regional accents, cultural backgrounds, and speaking styles
  • Fine-grained control over prosody, pacing, and tone
  • Supports spell() for spelling sequences letter by letter or number by number
  • Sinhala remains on Arcana after the sunset; all other Arcana languages migrate to Coda
  • Available with modelId: arcana

​
Mist v3

Mist v3, released March 2026, is the low-latency successor to Mist v2 and the preferred Mist model for new work.

​
Mist v2

For new work, prefer Mist v3: it offers lower latency and the same inline pronunciation control and custom pauses.
Mist v2, released February 2025, is the earlier low-latency Mist model.
  • English and Spanish support
  • Inline pronunciation control and custom pauses, also available in Mist v3
  • Approximately 70ms on-prem latency in Rime’s benchmark
  • 94 voices across accents, demographics, and speaking styles
  • Uses modelId: mistv2

​
Mist legacy

Mist, released April 2023, is a legacy model in the Mist family. Use modelId: mistv2 or modelId: mist to synthesize with these older deployments. As of February 2025, requests that omit modelId default to mist. Model v1 was released in April 2022 and has been deprecated.