Audio Quality Whitepaper
DNSMOS is the industry standard for scoring speech audio quality. It's also not built to tell you if your training data is right for your specific model.
Most real-world audio doesn't cleanly pass or fail. It lands in the grey zone between a DNSMOS of 3 and 4. Cut it, and you're discarding usable data. Keep it without calibration, and you risk quality regressions.
This whitepaper covers the calibration framework we use, how we validate audio quality before it ever reaches a model, and where DNSMOS holds up versus where it doesn't.
Written by Dheeraj Jalali (CTO), with Geoff Bremner (Senior AI/Datasets Engineer) and Justin Hugh (Product Manager, Voice Data).
Request access above to download the full whitepaper.
Once granted access, download it from the Files and versions tab above, or go directly to the PDF file.
License
Use of this whitepaper is governed by the Voices.com Whitepaper License.
- Downloads last month
- 1