Samples

Audio and speech demonstrations from my research.

  • Voice conversion samples
    Model output across three conversion tasks — speaker identity (timbre), accent, and speaking style — with source, target reference, and generated audio side by side for ten examples per task.
  • Cross-speaker pair corpora (LibriTTS)
    The same content spoken by two different speakers: content-matched word and phone pairs used to train diffusion voice conversion. Inline players across seven word and phone n-gram schemes.
  • Manx text-to-speech demo
    A short video of the Gaelg AI text-to-speech system reading Manx Gaelic aloud.