Verbissenso — European Portuguese lesson recordings

Prebuilt voices generated on 2026-11-06:
  pt-PT-RaquelNeural (Raquel, female)
  pt-PT-DuarteNeural (Duarte, male)

13,382 recordings: 6,691 exact-text MP3 files per voice. Generated using the
paid Azure Speech S0 resource Verbissenso-Voices in North Europe. All 14 jobs
succeeded with zero failed inputs. Provider usage: 505,922 neural characters.
The earlier F0 evaluation samples and Dii/TigreGotico recordings are not shipped.

Exact text is normalized to NFC and identified by the first 24 hexadecimal
digits of SHA-256. Voice releases use independent immutable directories.
azure-release.json contains the full decoding, non-silence, duration and SHA-256
verification manifest. azure-s0-provenance.json contains safe resource and job
provenance, with no API keys, tokens or temporary download URLs.
The approved native-speaker sample review does not certify every corpus clip.

Microsoft terms: https://www.microsoft.com/licensing/terms/en-US/product/MicrosoftAzure/MCA