Audio

Single male voice, mid-thirties, Filipino English accent. Room tone of a small office. No music, no processing.

Script:
"I told them the shipment would arrive Tuesday, and they just... [two second pause] ...they laughed at me. Sabi ko sa kanila, hindi ito biro. [breath] Anyway. It's fine. It's fine."

Direction: starts composed and professional, breaks into suppressed anger on "they laughed at me", then flattens into forced calm by the final line.

What each clause is testing

[two second pause]

Deliberate silence. Models compress or fill dead air because their training rewards continuous speech.

breaks into suppressed anger ... then flattens into forced calm

Two prosodic turns inside one unbroken take, with no line break to hide the transition.

Sabi ko sa kanila, hindi ito biro.

Code-switch into Tagalog mid-line. The accent and timbre must carry across without resetting to a default voice.

[breath]

Non-verbal generation. Most TTS is too clean to breathe unless told, and then breathes in the wrong place.

It's fine. It's fine.

Identical text, twice. A model with real prosody will not deliver them identically.

Room tone of a small office. No music, no processing.

Ambience control plus a negative instruction. Music beds are a common uninvited default.