Skip to content
← Models
ClosedCartesia/10 trustMetrics-checked

Sonic Turbo

Cartesia's speed-specialized TTS: the model behind the famous ~40ms time-to-first-audio claim — built for conversations where every millisecond counts.

Visit official page →

About

Sonic Turbo is Cartesia's latency-first text-to-speech model, the one behind the widely-quoted ~40ms time-to-first-audio figure. It's the speed member of the Sonic family: where Sonic 3 chases maximum quality and emotion in 40+ languages, Turbo trades some of that for the fastest possible first byte — the vendor's number refers to model inference, and third-party round-trip measurements typically land around 2x the claimed figure, which is still very fast. It covers 9 languages and plugs into Cartesia's platform features: a large voice library, high-fidelity voice cloning, SSML-style control tags (speed, volume, emotion — the latter in beta), and streaming over WebSocket. It's a staple of voice-agent stacks (Vapi, LiveKit, Home Assistant integrations all list it as a model option) precisely because it eliminates the awkward pauses that make agents feel robotic. Everything is closed and API-only, billed on Cartesia's usage plans, with SOC 2 Type II, HIPAA, and PCI compliance for regulated deployments.

Specs

Context window

Not applicable — text in, streamed audio out; 9 languages; model ID sonic-turbo

Pricing

Paid API (usage-based)

Free quota

Free trial credits on signup; then usage-based billing

API access

API available (REST + WebSocket streaming; Cartesia-Version header)

Weights

Closed — API only

Released

Not disclosed

Benchmarks

  • Provider-claimed ~40 ms time-to-first-audio (model inference only)
  • Third-party round-trip measurements (Vapi Humanness Index, Sonic-family lineage) land roughly 2x vendor claims
  • No independent benchmarks specific to Sonic Turbo found

Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.

Use cases

voice-agentsreal-time-communicationconversational-aiivr

Trust score — 6/10

/10 trustMetrics-checked

Scored 6/10 (TIER metrics, metrics-v1). Evidence: Cartesia's production platform (SOC 2 Type II, HIPAA, PCI); widely integrated by voice-agent frameworks (Vapi, LiveKit, Micdrop); current Sonic 3/3.5 generation. Strengths: genuinely fast streaming, voice cloning plus a large voice library, enterprise compliance and VPC/on-prem options. Docked for: (1) closed model — the headline 40ms figure is provider-claimed and third-party measurements run ~2x; (2) 9 languages trails the multilingual flagships, and quality trades against Sonic 3; (3) no public pricing transparency beyond usage-based plans. 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
  • · Model ID sonic-turbo: speed-specialized member of the Sonic family; provider-claimed ~40ms time-to-first-audio (model inference)
  • · 9 languages (third-party report); voice library + high-fidelity cloning; SSML-style tags (speed/volume/emotion in beta)
  • · Third-party round-trip measurements land ~2x vendor claims; no independent benchmarks specific to Sonic Turbo
  • · Widely integrated: Vapi, LiveKit, Micdrop, Home Assistant; SOC 2 Type II, HIPAA, PCI

Sources