Skip to content
SplendidBit

Last reviewed August 2, 2026

Third-party notices

FluidAudio

Vokyl uses FluidAudio 0.15.5 for on-device speaker diarization.

  • Copyright: FluidInference contributors
  • License: Apache License 2.0
  • Source revision: 19600a485baa4998812e4654b70d2bab8f2c9949

Speaker diarization Core ML models

FluidAudio downloads the FluidInference speaker-diarization-coreml model assets when speaker separation is first needed.

  • License: Creative Commons Attribution 4.0
  • Based on pyannote speaker segmentation, WeSpeaker embeddings, and VBx clustering
  • Model attribution and research citations are maintained in the linked model card

The model receives audio only through Core ML on the person's iPhone. The model host is contacted to retrieve model files; Vokyl does not upload recordings or transcripts.