About

Diatts is a weblog about open-source speech synthesis, written for people who build with these systems rather than for people shopping for one.

The subject here is text-to-speech as a working technology: how dialogue timing is modelled, how prosody is represented and controlled, how audio corpora are prepared, what a model card actually tells a reader, and how speech output can be evaluated without leaning on a single quality score.

What the notes are

Each note is a short technical essay on one of those questions. They are written to be read start to finish rather than skimmed, and they describe general practice rather than endorsing any particular implementation. Nothing published here ranks products, scores vendors or recommends a purchase.

Scope and voice

Notes are published under the name of the site. Where a note discusses consent, provenance or evaluation practice, it describes what a workable record or protocol contains; it is a description of common practice, not professional advice.

Finding things

The archive lists every note by month. The topics index groups the same notes under four subject headings.