That’s a common issue with TTS tone control usually tied to punctuation, SSML, and voice settings.
You can improve it by tweaking phrasing (clear question structure + ? ), adding SSML prosody (pitch/rate), or slightly adjusting the prompt so the model intends a question. ElevenLabs also responds well to small pauses or emphasis tags.
I’ve worked on voice agents like this I can help you fine-tune the tone to sound more natural and clearly question-like. Want to share a sample line or your current setup?
@Tamas Farago