Skip to content
← Models
ClosedPicovoice/10 trustMetrics-checked

Picovoice Cheetah

Streaming speech-to-text that never leaves the device — realtime transcription for apps and edge hardware.

Visit official page →

About

Cheetah is Picovoice's on-device streaming speech-to-text engine. All voice processing runs locally, which makes it a fit for privacy-sensitive apps and edge hardware: it runs on Linux, macOS, Windows, Android, iOS, web browsers and Raspberry Pi 3/4/5. The SDK bindings (Python, Node.js, .NET, Java, Go, Flutter, React Native, Web) are Apache 2.0 and a free Picovoice AccessKey gets you started, while the engine models themselves are proprietary. Picovoice publishes a vendor-run benchmark repo comparing word error rates against cloud providers; no independent reproduction was found, so treat the accuracy figures as vendor-measured.

Specs

Context window

Streaming audio frames; realtime

Pricing

Free tier with monthly transcription-hour allowance; paid plans for higher volume (see provider pricing page)

Free quota

Free tier with monthly transcription-hour allowance

API access

On-device SDKs (Python, Node.js, .NET, Java, Go, Flutter, React Native, Web); no cloud API — audio never leaves the device

Weights

Closed — API only

Released

Not disclosed

Benchmarks

Vendor-run benchmark repo (Picovoice/speech-to-text-benchmark) reports WER close to cloud services.

Independent: no third-party reproduction found — treat as provider-claimed.

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

Use cases

voice-assistantsedge-airealtime-captioningprivacy-sensitive-transcription

Trust score — 7/10

/10 trustMetrics-checked

Scored 7/10 (metrics tier, metrics-v1). Evidence: Picovoice/cheetah repo verified active (Apache-2.0 bindings, recent commits); cross-platform claims verified in the binding READMEs; pricing/free-tier terms documented by third-party coverage. Strengths: true on-device streaming, broad platform support, generous free tier. Docked for: (1) engine models proprietary and AccessKey-gated; (2) accuracy figures are vendor-run with no independent reproduction. License: SDKs Apache-2.0; models proprietary.

How the trust score is built

Full methodology →
Adoption · stars/downloads, log-scaled vs peers
7.0/10
Maintenance · recency, releases, issue load
9.0/10
Openness · license + docs
5.0/10
Manual review · sanity read + red-flag check
8.0/10
  • · 670 stars on Picovoice/cheetah; actively maintained
  • · Fully on-device streaming STT: Linux/macOS/Windows, Android/iOS, browsers, Raspberry Pi
  • · WER claims are vendor-run (Picovoice/speech-to-text-benchmark)
  • · Free tier with monthly hour allowance

Sources