Arcana
Rime's most expressive TTS: trained on annotated everyday conversation — hesitations, laughter, and <laugh> tags included — now serving Arcana v3.
About
Arcana is Rime Labs' flagship expressive text-to-speech model, built on an unusual bet: that the best training data isn't studio recordings but messy, everyday conversation, richly annotated with its context. The result is a voice model that handles the paralinguistic texture of real speech — laughter, sighs, hesitations, overlaps — with native <laugh>-style tags and dynamically-generated custom voices, plus unlimited text input, strong text normalization, and MP3/WAV/PCM/mu-law output. Rime's stack powers 100M+ calls a month for customers in healthcare, finance, and telecom, and the company raised a $24M Series A in July 2026 to build enterprise speech-to-speech. The model you get today is Arcana v3 (announced February 2026, auto-served on the same arcana model ID): Rime says it folds Mist v2's speed into Arcana's richer tone, and it's also available on Together AI's dedicated infrastructure. Rime is enterprise-first — SOC 2 Type II and HIPAA, cloud/VPC/on-prem deployment — and the API streams over HTTP, WebSocket, and SSE. No public benchmarks exist; the quality evidence is partner integrations (SignalWire, Together AI) and Rime's own customer claims.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
- No published benchmarks
- Partner listings (SignalWire, Together AI) position it as a flagship conversational voice model (partner claims, not independent benchmarks)
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (TIER metrics, metrics-v1). Evidence: Rime Labs' production platform (founded 2022; $24M Series A July 2026; 100M+ monthly calls); Arcana v3 announced Feb 2026 with Together AI dedicated hosting; SOC 2 Type II + HIPAA. Strengths: distinctive conversational-speech training approach, native paralinguistic tags, custom voices, enterprise deployment options (cloud/VPC/on-prem). Docked for: (1) closed model with zero published benchmarks — all quality evidence is partner and vendor claims; (2) enterprise positioning means pricing and free-tier details are opaque; (3) no real brand logo obtainable for the catalog — omitting logoUrl. License: proprietary.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.0/10
- Maintenance · recency, releases, issue load
- 8.0/10
- Openness · license + docs
- 2.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · Announced May 14, 2025; Arcana v3 auto-served on the arcana model ID since Feb 2026
- · Trained on annotated everyday conversational speech; native paralinguistic tags (<laugh>), custom voices, unlimited text input
- · Rime: $24M Series A July 2026, 100M+ calls/month; SOC 2 Type II + HIPAA; cloud/VPC/on-prem
- · Zero published benchmarks — quality evidence is partner/vendor claims only; no real brand logo obtainable, logoUrl omitted