Wake word and voice command data, recorded to your spec
We record your wake phrase, the commands after it and the phrases that sound like it, with native speakers recruited to your accent and demographic quotas. Sessions are recorded on the devices and in the environments your product will meet, from phones and headsets to far-field and in-car audio, and every one is checked before delivery.
Positives, negatives and the voices that say them
Your phrase list
Speakers record the wake word, the commands that follow it and the confusable phrases that should not trigger it, from the prompt list in your brief.
Speakers who sound like your users
Native speakers are recruited to your quotas for age, gender, region and accent. Each passes an ID check and a microphone test, and we confirm their language or dialect.
Recorded where your product is used
Sessions are recorded on the devices and in the environments your product will meet, then go through automatic checks for noise, clipping and duration and a human review.
What a project can include
| Prompts | Wake word, commands and confusable phrases from your list |
|---|---|
| Speakers | Native speakers recruited to your quotas for age, gender, region and accent |
| Devices | Phone, headset, far-field or in-car, as your brief specifies |
| Transcripts | Time-aligned, human-verified transcripts in your convention, with extra labels as your spec requires |
| QA | Automatic checks plus human review; sessions that miss spec are re-recorded, not shipped |
| Consent | Explicit commercial-use consent from every speaker, documented |
| Languages | Any of our 60 catalog languages and dialects, and others after a feasibility check |
Built for
- Wake word and keyword spotting models
- Voice assistants and smart speakers
- In-car voice control
- On-device command recognition
- False-accept testing with confusable phrases
Catalog or custom?
Recordings of your wake phrase are always made to spec. For general speech to use as negatives, our off-the-shelf datasets hold spontaneous conversation in 60 languages and dialects. For how detectors are trained and measured, read about wake word detection.
Frequently asked questions
Where can I get wake word training data recorded in my target languages?
SpeechData.ai records wake word and voice command data to order. Native speakers recruited to your accent and demographic quotas record your wake phrase, commands and confusable phrases on the devices you specify, and every session is checked before delivery. You get a feasibility assessment and a quote within one business day of sending your brief.
Can you record confusable phrases as well as the wake word?
Yes. Speakers record whatever your prompt list contains: the wake word, the commands that follow it, and phrases that sound like it but should not trigger the detector. They are recorded and transcribed the same way, so you can train and measure false accepts.
Which languages can you record?
Any of the 60 languages and dialects in our catalog, and other languages on request. We confirm recruitment feasibility for each language before quoting.
Describe your wake word and your users
Share your phrases, languages and accents, speaker counts, devices and environments. We’ll come back within one business day with a feasibility assessment and a quote.