Questions Regarding Voiceflow Chat Widget Voice Behavior & Pronunciation
I'm currently using the Voiceflow web chat widget with voice output enabled, and we’ve encountered a few issues we’d like your help with:
1. Voice Reads Text Verbatim from Chat
We've customized our LLM prompts so that the chat responses are much more user-friendly and concise. However, the voice still reads the full chat response word-for-word, even when it's long or pulled directly from website content.
👉 Is there any way to customize the spoken output separately from the text shown in chat (e.g., make it more summarized or conversational)?
2. Manual Mic Input – Not Natural
Currently, users have to click the mic button, speak, and then press Enter to send their message. This feels like basic voice-to-text, not a seamless voice assistant experience.
👉 Is there a way to automate message submission after voice input, or support continuous voice conversation flow?
3. URL Reading in Voice
The assistant currently reads out full URLs, which sounds awkward. Ideally, we want URLs to appear in the chat but not be spoken aloud.
👉 Is there a way to suppress URL pronunciation in voice output?
4. Pronunciation Issues with Abbreviations (e.g., “CM”, “GSM”, “L”)
We're facing issues where the assistant mispronounces certain units and abbreviations, like “CM”, “GSM”, or “L” for litres.
👉 Is there any support for custom pronunciation dictionaries or phoneme mapping within the chat widget voice settings?
0 Replies