Voice replies start sooner after a period
October 16, 2026 · platform
The voice engine reads a period followed by a lowercase word as an abbreviation, like "approx. eight". A reply such as "Okay. eight one six" was therefore synthesized as one sentence, and its first audio started later. On our speech servers the first audio chunk took about 250 ms instead of about 100 ms.
Agents now fix this on the way to the engine. The first letter after ". ", "! " or "? " is uppercased in the copy of the text handed to synthesis, so the engine reads the period as a sentence end and the reply starts sooner.
Your text stays as it is
Only the copy sent to the voice engine changes. Your speak() text, speech events, transcripts, conversation history and agent logs keep the text exactly as you sent it. Speech has no letter case, so the words sound the same. Ellipses, abbreviations like "e.g." and "Dr.", and decimals such as 1.5 are left alone.
Turning it off
The setting is on by default. If you rely on lowercase after a period, for run-on text or for notation where a capital changes the meaning, switch it off and deploy again.
- Docs: Sentence start capitalization
- Dashboard: Advanced voice pipeline, then TTS playback, then Sentence start capitalization
- CLI:
voicethere projects voice-advanced set tts.sentenceCase off