Thanks — this is a great use case. I’m responding to your question about whether LiveKit can showcase multiple Gemini TTS voice “personalities” (calm, energetic, professional) and how to do it without disrupting the main user flow. Short answer • Yes. You can use Gemini TTS with LiveKit to present different voice personalities by sending descriptive

instructions

(prompting) and switching `voice_name`/options for each selection. • Do the change on the TTS/options layer (e.g., with an update call) so it only affects future utterances and doesn’t interrupt the current session. How this works (what to change) • Personality = descriptive instructions passed to the Gemini TTS engine (tone, pacing, emotion). ◦ Example prompts: ▪︎ Calm: "Speak in a calm and soothing tone, with slow pacing and gentle intonation." ▪︎ Energetic: "Speak in an energetic and enthusiastic tone, with lively pacing and dynamic expression." ▪︎ Professional: "Speak in a clear, confident, and professional tone, with moderate pacing and precise articulation." • Voice selection = choose a Gemini voice name (e.g.,

Zephyr

,

Kore

) and pair it with the instructions to shape personality. • Apply changes using the TTS update method so the next utterance uses the new personality/voice without restarting the agent or interrupting the conversation. Implementation approach (recommended) 1. Maintain a small voice-state per user/session that stores: ◦ Selected

voice_name

◦

instructions

(the personality prompt) 1. Provide a UI control (buttons or a dropdown) or a tool trigger to let the user pick a personality. 2. When a user picks a personality: ◦ Update the session’s TTS options (so only subsequent TTS output changes). ◦ Do not reinitialize the whole session/agent—use the update API. 1. Ensure updates take effect on the next utterance: ◦ This avoids cutting off or affecting in-progress audio. 1. Optionally add a short preview button that plays a single sample utterance using the selected personality so users can compare before committing. Concrete examples • Python (conceptual)

# Update TTS options for the current session so next utterance uses new personality
session.tts.update_options(
    voice_name="Zephyr",
    instructions="Speak in a calm and soothing tone, with slow pacing and gentle intonation."
)

• Node/TypeScript (conceptual)

session.tts.updateOptions({
  voiceName: "Zephyr",
  instructions: "Speak in an energetic and enthusiastic tone, with lively pacing and dynamic expression."
});

UX tips to avoid impacting main flow • Apply changes only on user action (explicit selection or preview), not automatically during active speech. • Keep a short confirmation or quick sample playback so users know what changed. • If your app has critical real-time voice interactions, only allow personality switches in non-critical moments or at natural breakpoints (e.g., between messages). • Cache preference per user so repeated selections don’t cause repeated updates. Limits / notes • Use

instructions

to shape personality — that’s the mechanism Gemini TTS exposes for tone/pacing. • Voice changes via update methods apply to the next utterance; they won’t retroactively change already-played audio. If you’d like, I can: • Draft ready-to-use personality prompts tuned for your use case (short, mid, and long variants). • Provide a minimal UI flow example that shows selection → update → preview without interrupting audio. • Help you pick default Gemini voices and map them to each personality (calm/energetic/professional). Which of these would help you next? And are you building this into an existing LiveKit room/agent session or starting a new demo flow? Sources: Gemini TTS plugin guide | LiveKit Documentation