Vosk
Offline speech recognition in 20+ languages that runs on a Raspberry Pi and needs no cloud.
About
Vosk is Alpha Cephei's offline speech recognition toolkit, built for places the cloud cannot reach. Per-language models weigh about 50 MB, run on a Raspberry Pi or a mid-range phone, and support more than 20 languages and dialects, from English and Mandarin to Kazakh and Esperanto.
It ships a streaming API with near-zero perceived latency, word-level timings and confidence scores in structured JSON output, plus vocabulary reconfiguration and speaker identification. Bindings cover Python, Java, Node.js, C#, C++, Rust, and Go, and the whole thing is Apache-2.0. Accuracy will not match server-class models, but for chatbots, smart-home appliances, and lecture transcription that must work offline, it remains the pragmatic choice.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
No published standard benchmark headlines; ~50 MB models trade some accuracy for size and speed
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 (verified tier, metrics-v1). Evidence: Apache-2.0 license, ~14,800 GitHub stars, and 20+ supported languages confirmed across the repo and independent coverage; ~50 MB per-language model size and streaming JSON API from the project docs. Strengths: the longest-running offline ASR toolkit still maintained (commits into 2026), genuinely tiny models, broadest language binding coverage. Docked for: (1) no published benchmark figures, so accuracy claims rest on community reputation; (2) small models visibly lag server-class accuracy on hard audio. License: Apache-2.0.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 8.0/10
- Maintenance · recency, releases, issue load
- 7.0/10
- Openness · license + docs
- 10.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · ~14,800 GitHub stars; 20+ languages/dialects; ~50 MB per-language models
- · Streaming JSON API with word timings + confidence; Python/Java/Node/C#/C++/Rust/Go bindings