Voice recording · Wake Word Detection
Voice Recording for AI Training for Wake Word Detection
Studio voice recording built for model training rather than broadcast: controlled acoustics, consistent mic distance, and reproducible session parameters across every speaker. Applied to wake word detection, the specification is driven by one thing: false accepts per hour.
- Service
- Voice recording
- Use case
- Wake Word Detection
- Primary metric
- False accepts per hour
Required data profile
- Thousands of speakers, few utterances each
- Positive and hard-negative sets
- Multiple distances and noise conditions
Technical specification
| Parameter | Standard |
|---|---|
| Sample rate | 48 kHz |
| Bit depth | 24-bit |
| Microphone | Large-diaphragm condenser, fixed distance, pop filter |
| Room | Treated booth, RT60 under 0.3s |
| Processing | None. No compression, EQ, or noise reduction on delivered audio |
| Session length | 30-90 minutes per speaker, with breaks logged |
Process
- Session template definition so every studio in the network records identically
- Speaker briefing and pronunciation calibration
- Take-level monitoring with immediate re-record on defect
- Per-session technical report
- Delivery with unprocessed masters
Metrics this feeds
- False accepts per hour
- False reject rate per accent band
- Performance at 3m and 5m
Failure modes to design out
- Positives only, with no hard negatives
- Close-mic-only capture
- No accent-band tagging, so failures cannot be localised
Delivered audio is never processed. Training pipelines should see the raw capture so augmentation stays under your control.
Deliverables
- Unprocessed WAV masters
- Session logs
- Take-level metadata
- Speaker consent records
Frequently asked
Is voice recording the right service for wake word detection?
It covers thousands of speakers, few utterances each. Most wake word detection programmes combine it with at least one other service; we will say so in the scope rather than selling one line item.
What languages are available?
All 14 languages in the network plus Indian English accent bands.
How is the evaluation set handled?
Collected first, from speakers disjoint from the training cohort, so improvement is measurable.
Scope voice recording for wake word detection
Send the metric you need to move.