# Picovoice/porcupine

On-device wake word detection powered by deep learning

Repository: https://github.com/Picovoice/porcupine
Canonical: https://ross.abutalabs.com/products/porcupine
Homepage: https://picovoice.ai/
Language: Python
License: Apache-2.0
License Family: permissive
Topics: wake-word-detection, hotword, keyword-spotting, keyword-spotter, wake-word, wake-word-engine, handsfree, hotword-detection, hotword-detector, on-device, speech-recognition, trigger-word-detection, voice-activation
Last push: 2026-08-12T19:22:06+00:00

## Health v2 (maintenance only)
Score: 73/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 97, release rhythm 28, longevity 100
- inputs: {"age_days": 3101, "days_push": 21, "days_rel": 265, "gap_med": null, "n_releases_24m": 1}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 4922, forks 577 (observed 2026-08-28T04:09:02.439365+00:00)

## What it is
Porcupine is a highly-accurate, lightweight on-device wake word detection engine powered by deep neural networks. It enables always-listening voice-enabled applications across mobile, web, desktop, and embedded platforms, with custom wake word training via Picovoice Console.

## Use cases
- detect a wake word to activate a voice assistant
- add hands-free voice activation to an IoT device
- run keyword spotting entirely on-device without cloud calls
- trigger voice commands on a Raspberry Pi or microcontroller
- add wake word detection to an Android or iOS app
- listen for a custom hotword in a browser-based app

## When to choose
- you need privacy-preserving, fully on-device wake word detection
- you target embedded or IoT hardware with limited compute
- you need cross-platform SDKs (Python, Node, Android, iOS, web, embedded)
- you want to train custom wake words with a self-service console

## When to avoid
- you need full speech-to-text transcription rather than keyword spotting
- you require a fully open-source model pipeline (custom models go through Picovoice Console)
- you need offline training of wake word models without external services

## Facets
- artifact type: library
- maturity: stable
- function: speech-recognition, machine-learning, audio-processing, sdk
- domain: speech-processing, machine-learning, iot, embedded-systems, cross-platform
- platform: cross-platform, python, embedded, windows, browser, wasm
- tags: wake-word-detection, keyword-spotting, hotword-detection, voice-activation, on-device, always-listening, handsfree, nodejs, android, ios, web-server, macos, linux

## Member repositories
- Picovoice/porcupine (main) score 73

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:02.439365+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T18:18:01.447809+00:00, confidence not recorded.
  - readme: https://github.com/Picovoice/porcupine (fetched 2026-08-28T04:09:02.439365+00:00, sha d47cb4bb6df0)
  - homepage: https://picovoice.ai/ (fetched 2026-08-29T08:59:55.196515+00:00, sha 44136fa355b3)
- Data as of 2026-08-30T08:39:29.467469+00:00.