Docs Book a Demo Sign in
Sign in Start building

Credits

Last updated: November 1, 2026

Lokutor builds its own speech models. Some features also use open-source work by others, and we credit it here.

Speaker diarization

The speaker labels in our speech-to-text, which say who spoke each word, use the pyannote/speaker-diarization-community-1 pipeline from pyannote.audio, licensed under the Creative Commons Attribution 4.0 International licence (CC-BY-4.0).

Reference voices

Each of our ten voices is a short reference recording of a real person who read text aloud for an open speech corpus. We credit the corpora here; the licence of each is the one its authors published. The speakers are corpus readers, not product voice actors.

  • English: VCTK (University of Edinburgh); Hi-Fi TTS (NVIDIA). CC BY 4.0.
  • Spanish (Spain): Mozilla Common Voice 17, Spanish. CC0.
  • German: Multilingual LibriSpeech (Meta); HUI-Audio-Corpus-German (LibriVox readers). CC BY 4.0; CC0.
  • French: SIWIS French Speech Synthesis Database (University of Edinburgh). CC BY 4.0.
  • Portuguese: Mozilla Common Voice, Portuguese. CC0.
  • Italian: Mozilla Common Voice 17, Italian. CC0.
  • Catalan: Festcat (Universitat Politècnica de Catalunya, Barcelona Supercomputing Center); Crowdsourced Catalan speech (OpenSLR 69, Google). CC BY-SA 4.0.
  • Galician: CRPIH UVigo GL Voices (Universidade de Vigo); Crowdsourced Galician speech (OpenSLR 77, Google). CC BY 4.0; CC BY-SA 4.0.
  • Basque: Crowdsourced Basque speech (OpenSLR 76, Google). CC BY-SA 4.0.

Where a licence is share-alike (CC BY-SA 4.0), the speaker's recording is credited here. The same voice names stay across versions, but the person behind a voice in a language can change when we replace a reference with a clearer one. The docs have a list too: reference voices in the docs.