Fast, accurate, and lightweight custom wake word detection
Train custom wake words in seconds, deploy across embedded, mobile, web, desktop, or server. Enterprise-ready. No training data required.
Train and test “Hot Pink” or a custom wake word by typing it.
or click mic to try another phrase
The only wake word engine you'll ever need
Porcupine Wake Word is an on-device keyword spotting (KWS) engine that enables "always-on" voice interfaces. It listens for a specific keyword or phrase to trigger voice-enabled applications.
Porcupine Wake Word is what a wake word detection engine should be: lightweight, accurate, customizable, and production-ready. With Porcupine Wake Word, enterprises can train branded wake words and always-listening commands in seconds and deploy them across embedded, web, mobile, and desktop, with all inference fully on-device.
No training data, machine learning pipeline, or special infrastructure is required, making wake word integration fast, reliable, and resource-efficient.
Each day, someone is in danger in unmonitored areas. They're attacked, threatened, or experience a medical emergency and need someone to hear the call for help. With HALO, using Picovoice technology to recognize keywords, security personnel can now respond to the call.
Simple wake word integration in any application
Integrate wake word detection into your app with just a few lines of code. Porcupine Wake Word provides SDKs for Python, NodeJS, Android, iOS, React, Flutter, React Native, .NET, Java, C, and Web, enabling rapid deployment across embedded, mobile, web, desktop, and server.
Proven Accuracy and Efficiency
When comparing wake word engines, accuracy must be balanced with power consumption. Porcupine Wake Word leads the market in both.
Why enterprises choose Porcupine Wake Word
Porcupine is an enterprise-ready on-device wake word engine built for high accuracy, low resource usage, and ease of integration. It runs always-on across platforms, supports flexible custom training, and is private by design.
Ship it.
On device.
Fast, accurate, and lightweight wake word detection
Common questions about wake word detection
A wake word is a unique phrase that activates dormant applications. For example, Amazon, Apple, and Google devices wake up when they detect Alexa, Hey Siri, and OK Google. Wake word, trigger word, hotword, and wake-up word are used interchangeably.
Wake Word Detection is one of the applications of Keyword Spotting (KWS) technology. It detects (spots) phrases (keywords) in audio streams and conversations. Voice activation is the most common use case for wake word detection.
Trigger Word Detection, Hotword Detection, and Wake Word Detection are interchangeable terms. For example, NASA uses the terms hot word recognition in one project and wake word detection in another, despite using the same product: Porcupine Wake Word Detection. Google predominantly uses hotword detection but doesn't offer custom hotwords.
A wake word detection engine is a binary classifier that recognizes pre-defined phrases. During training, the detection engine learns the desired wake word and how to differentiate it, so when integrated into software listens to the environment to detect that keyword.
- be accurate with low FAR and FRR,
- run locally on the device to achieve affordable, power-efficient, and private hands-free experiences,
- be resource-efficient, and fast, as wake word detection engines listen to conversations in the environment. That's why using other solutions, such as automatic speech recognition for wake word detection, is not a good idea.
To learn more about wake word detection requirements, we suggest our complete guide to wake word detection.
The performance of the wake word depends on several factors, including the number of phonemes, vowels, and syllables. The best wake words are six or more phonemes long, phonetically distinct from common speech, and contain a mix of vowel sounds. Avoiding names, common commands, or phrases that resemble everyday conversation is recommended as they lead to unintended activations. For more information, check out the guide on choosing a wake word.
Picovoice doesn't gather or require customer data, thanks to transfer and self-supervised learning algorithms in Porcupine Wake Word.
No. Training a branded wake word on the self-service Picovoice Console does not even require any coding skills. You can simply type your desired wake word and get a working model in seconds. To learn how to train a custom wake word in seconds, visit the Porcupine Wake Word docs and watch or read the custom wake word training tutorial.
Porcupine detects a single wake word to activate your application. Rhino is Picovoice's Speech-to-Intent engine — it understands natural language commands after activation. They are commonly used together: Porcupine listens continuously for the wake word, then passes control to Rhino to understand what the user wants to do.
- Using Wake Word to Build Voice Activation Applications:
- Smart home devices and IoT applications
- Mobile app voice controls
- Automotive infotainment systems
- Healthcare and accessibility tools
- Industrial automation and robotics
- Using Wake Word to Build Always-Listening Commands:
- "Turn the lights on" for home automation
- "Start recording" for productivity apps
- "Emergency help" for safety applications
- Custom brand-specific activation phrases
- Using Wake Word to Build Keyword Monitoring:
- Brand mention detection in conversations
- Profanity filtering in content moderation
- Compliance monitoring for regulated industries
- Product name recognition in customer service
Picovoice docs, blog, Medium posts, and GitHub are great resources to learn about voice AI, Picovoice technology, and how to start using wake words. Enterprise customers get dedicated support specific to their applications from Picovoice Product & Engineering teams. Reach out to your Picovoice contact or talk to sales to discuss support options.
Porcupine Wake Word empowers developers to train any wake word of choice and always-listening custom commands that could work with Alexa-enabled applications and Google Assistant. Technically, "Jarvis" or other phrases replace "Alexa" and "Hey Google". In practice, Amazon and Google policies determine what developers can use.
Training a custom wake word on Picovoice Console takes less than ten seconds, visit Porcupine Wake Word docs, read custom wake word training tutorial, or watch it. After downloading your platform-optimized wake word, integrate it into your product with one of the Porcupine Wake Word SDKs.
Comparing hotword, trigger word, wake word, or wake-up word models the right way is complex. Learn more about the terms, such as FAR, FRR, and ROC, used in evaluations and use the open-source benchmark whether your vendor calls it hotword, trigger word, wake word, or wake-up word.
Porcupine Wake Word supports English, French, German, Italian, Japanese, Korean, Chinese (Mandarin), Portuguese, and Spanish.
Contact sales to get a custom wake word model trained in any language for your application.