Picovoice Cheetah

by Picovoice

On-device streaming speech-to-text optimized for embedded and edge.

TL;DR

On-device streaming speech-to-text optimized for embedded and edge.

Best for on-device streaming STT for voice assistants, wearables, hearing aids, kiosks. Pricing: Free tier + per-user/Enterprise plans.

Category
Products
License
—
Stars
—
Last push
—
Pricing
Free tier + per-user/Enterprise plans
Platforms
Edge, SDK, iOS, Android, Linux, Windows, macOS

What it is

Picovoice Cheetah is an on-device streaming speech-to-text engine designed to run within tight CPU/RAM budgets, including microcontroller-class hardware. It supports incremental partial transcripts, word timestamps, and a fixed inventory of languages. Picovoice's other relevant products include Leopard (offline batch STT), Porcupine (wake word) and Falcon (speaker diarization). Pricing is access-key-based with a free non-commercial tier and paid enterprise plans. Best fit: on-device streaming stt for voice assistants, wearables, hearing aids, kiosks. Caveats: vocabulary and language coverage smaller than cloud apis; requires picovoice access key. Pricing as listed: Free tier + per-user/Enterprise plans. Feature flags from vendor docs: word-level timestamps, streaming. Directory tags: embedded, on-device. Last vendor-page check: 2026-05-12.

Best for: On-device streaming STT for voice assistants, wearables, hearing aids, kiosks.
Watch out for: Vocabulary and language coverage smaller than cloud APIs; requires Picovoice access key.

Install / use

pip install pvcheetah  # or platform SDK

Features

Speaker diarizationNo
Word-level timestampsYes
Streaming / real-timeYes
Languages supported7
HIPAA eligibleNo

Picovoice Cheetah vs Whipscribe

FeaturePicovoice CheetahWhipscribe
CategoryProductsTranscription APIs
PricingFree tier + per-user/Enterprise plansfree beta
Speaker diarizationNoYes
Word timestampsYesYes
StreamingYesNo
Languages799
PlatformsEdge, SDK, iOS, Android, Linux, Windows, macOSWeb, API, MCP

Alternatives to Picovoice Cheetah

Otter.ai
Otter.ai

Meeting-bot transcription product for Zoom/Meet/Teams.

free / from $10/mo
Rev
Rev

Human + AI transcription, highest accuracy tier on the market.

AI from $0.25/min · human from $1.50/min
Descript
Descript

Audio/video editor that treats the transcript as the timeline — different product category.

free / from $12/mo

Whipscribe is a managed faster-whisper + whisperX service. If you want transcripts without running infrastructure, paste a URL or drop a file in the form below — you'll have a transcript in seconds.