Wake Word Detection

Picovoice's wake word detection software enables building products that can be activated and controlled with voice. It makes it possible to activate the device similar to “Alexa” or “OK Google” but using your word word of choice. Additionally, the library allows the user to control the device via a configurable set of voice commands.

Accurate

Accurate

Outperforms existing solutions with high margins in clean and noisy environments [1].

Lightweight

Lightweight

The natural fit for IoT and memory/compute constrained applications.

Multiple Hotwords

Multiple Hotwords

Can detect multiple (many) wake words with no additional CPU or memory footprint.

Zero Lead Time

Zero Lead Time

Uses proprietary AI algorithms to build models for any custom wake word within seconds (instead of weeks). Removes the hassle of time-consuming and costly data gathering phase.

Always On

Always On

Optimized and power efficient implementation designed for always listening applications.

Ease of Integration

Ease of Integration

Expedite your development by reusing the many available reference designs.

Interactive Demo

This demo allows you to change the color of the smart lamp using Picovoice's wake word engine via any of the following commands. Note that you need a working microphone. This demo is running locally in the browser. You can turn off your internet connection and it will keep working.

  • OK lamp, white.
  • OK lamp, yellow.
  • OK lamp, orange.
  • OK lamp, purple.
  • OK lamp, navy blue.

Start Demo

Light Bulb

Speech to Intent

The Speech to Intent engine infers user's intent from complex voice commands. It recognizes variations in spoken language to provide a natural voice interface. It can run on even tiny processors such as ARM Cortex-M or fixed-point DSPs.

Language Understanding

Language Understanding

Directly infers user's intent from speech. Improves user experience.

Lightweight

Lightweight

The natural fit for IoT applications. Can run on almost any embedded/mobile platform.

Customizable

Customizable

Can be readily customized (optimized) for your product and use cases.

Realtime

Realtime

Infers intent from speech in real time. All processing is done on device with zero latency.

Always On

Always On

Can be paired with wake word engine to run in always-on mode with no additional footprint.

Multiple Commands

Multiple Commands

Can detect many commands (intents) with no additional CPU or memory footprint.

Contact Us View on Github

Interactive Demo

This demo allows you to issue voice commands to a smart coffee maker. Note that you need a working microphone. You can ask the coffee maker (Alfred) to make you a caffeinated drink using voice commands such as

  • Alfred, can I have a latte?
  • Alfred, make me a single-shot espresso.
  • Alfred, I want a triple-shot americano with milk.
  • Alfred, may I have a large cappuccino with cream?
  • ...

This demo is running locally in the browser. You can turn off your internet connection and it will keep working.

Start Demo

coffee maker

Small   Medium   Large  

Single Shot   Double Shot   Triple Shot  

Americano   Cappuccino   Espresso   Latte   Mocha  

Milk   Cream  

Speech to Text

Picovoice's Speech to Text software transcribes audio to text without the need to an internet connection. It is similar to speech recognition offerings from Google or Amazon cloud services but all processing is done locally on device.

Accurate

Accurate

Achieves 90% word accuracy. Outperforms existing solutions [1].

Realtime

Realtime

Transcribes audio in real-time. All processing is done on device with zero latency.

Lightweight

Lightweight

The natural fit for IoT. Runs on small embedded platforms such as Raspberry Pi Zero.

Open Vocabulary

Open Vocabulary

No limit on the size of vocabulary. Can transcribe any number of words.

Customizable

Customizable

Build extensible and context-aware models for your domain of interest.

Compressed

Compressed

Uses a proprietary model compression algorithm to reduce the size of SDK to as low as few MBs.

Contact Us View on Github View Benchmark