Primitive of the day · 2026-10-14

dia-parenthesized-nonverbal-whitelist

Dia non-verbal cues are PARENTHESIZED — (laughs), never [laughs] — and only the 21 tags listed in the Dia README are recognized; even those may give unexpected output, so use them sparingly.

When this applies

Adding laughter/sighs/coughs etc. to a Dia transcript.

Preconditions

Do this

Normalize [tag] -> (tag) for listed tags; warn on any unlisted parenthesized tag. The listed tags (Dia README, checked 2026-10-23): (laughs), (clears throat), (sighs), (gasps), (coughs), (singing), (sings), (mumbles), (beep), (groans), (sniffs), (claps), (screams), (inhales), (exhales), (applause), (burps), (humming), (sneezes), (chuckle), (whistles).

What you should see

Non-verbals render as sounds, not spoken words or artifacts.

How it fails if ignored

Square brackets are Dia's speaker tags ([S1], [S2]), not the non-verbal syntax; unlisted or overused non-verbals may cause odd audio artifacts, which only listening to the output reveals.

Do not use when

TTS engines with different cue syntaxes; SSML-based systems.

Corrected on 2026-10-23: the 21 tags are now listed here, as in the Dia README, and an internal remark was removed.

Kind: gotcha-fix. Part of the skill Dia dialogue TTS. Free to reuse in your own agent skills.