🗣️ Bozos TTS Guide

When TTS sounds wrong: check language, locale, voice and speaker mode in order

Bozos has several independent layers: voice-channel or personal language settings, regional locales, neural voices, pronunciation processing and optional Spatial Speakers. Change one layer at a time.

Start with the language and regional locale

Run /language and confirm the selected language matches the text being spoken. Regional locales can change rhythm, accent and pronunciation, so choose the correct locale before changing voices.

Understand Voice channel vs Only me

The current bot does not use a single global server-wide TTS default for these commands. Voice channel changes the default for the current VC. Only me creates your personal override. If your messages sound different from someone else's, this may be expected.

Try another compatible voice

The production catalog contains 323 verified voices. Use /voice to choose another voice in the same locale without changing the language/accent itself.

Use /voicepreview before committing

/voicepreview privately lists and previews voices for your currently effective language/accent. It is the fastest way to compare voices without sending test speech through the VC.

Check the active speaker system

If Spatial Speakers is active, Bozos can automatically assign compatible voices and spatial positions to members. A personal /voice selection still overrides the automatically assigned Spatial Speakers voice. Use /speaker-system map to see the current assignments.

If you want conventional spoken usernames again, switch the current VC back to the Username System with /speaker-system set.

Pronunciation processing is separate

The Grand Pronunciation Engine improves slang, acronyms, technology/gaming terms, file formats, units and Discord-style names. If staff temporarily disable the pronunciation feature globally, normal TTS can continue while those extra corrections are bypassed.

Numbers, URLs, emoji and formatting

Bozos normalizes many chat-oriented text patterns before synthesis. If one unusual message sounds wrong, test a plain sentence in the same language and voice. If plain text is correct, the issue is likely the specific formatting or term rather than the selected voice.

Fast troubleshooting order

  1. Confirm Voice channel vs Only me scope.
  2. Confirm the intended language.
  3. Confirm the regional locale/accent.
  4. Run /voicepreview.
  5. Try another voice in the same locale.
  6. Check whether Spatial Speakers is active and inspect /speaker-system map.
  7. Test a simple sentence.
  8. Only then investigate a specific pronunciation edge case.
Bozos TTS documentation →TTS setup guide →