Set Up Gemini Auto-Tone Mode
Auto-Tone Mode is a Gemini-only TTS feature that uses Gemini Flash to analyze a donation message and pick a voice + style prompt that fits its mood automatically — e.g. a funny donation gets a cheerful voice, a "send me good vibes" donation gets a warm one.
The result: TTS that sounds way more natural and emotionally aware — like an actual person reading.
What you'll need
- Your Gemini API key connected in Settings → Connections
- Gemini selected as the Primary TTS provider (see the TTS setup guide)
- An understanding that this feature adds ~500ms of latency per message (because Gemini Flash runs first, then the audio is generated)
Step 1 — Confirm Gemini is your Primary Provider
- Go to Settings → TTS Voice
- Check that Provider = Gemini
- If not — select Gemini first
Step 2 — Enable Auto Mode
- In the Gemini config section, find the Auto Mode toggle (✨)
- Flip it to ON
- The manual voice picker is now overridden by the AI
Step 3 — Pick an Emotion Intensity
With Auto Mode on, an Emotion Intensity dropdown / slider appears. Pick one of 4 levels:
| Level | Feel |
|---|---|
| Subtle | Light tone shifts — good for chill / lo-fi streams |
| Moderate | Clear tone changes without overdoing it (recommended default) |
| Strong | Big emotional contrast — fun / hyped / sad clearly distinct |
| Dramatic | Voice-actor levels of expression — perfect for variety / hype streams |
Step 4 — Test with Preview
- Try a few messages with different vibes, e.g.:
- "Thank you so much, I'm super happy!"
- "I'm feeling kinda down — could use some good vibes"
- "Today was the best stream ever!"
- Hit Preview and listen — does the AI pick noticeably different voices / tones?
- Bump Intensity up or down if the result feels flat or over the top
Cost & Latency notes
:::warning Slight extra cost Auto-Tone calls Gemini Flash for every message before calling Gemini TTS — that's a small extra cost per donation (Flash is very cheap, but not free). Track your usage in Google AI Studio. :::
:::caution +500ms latency Each donation is roughly half a second slower than manual mode. On streams with rapid-fire donations you might feel it — drop to Subtle or turn Auto Mode off during peaks. :::
Tips & Troubleshooting
- AI keeps picking the same voice: bump Intensity up to Strong or Dramatic — Subtle gives less variety by design.
- Auto Mode not working: double-check your Gemini API key in Connections and that Provider is actually Gemini.
- Want to override per-message: Auto Mode is all-or-nothing — turn it off and write your own style prompt if you need fine control.
- Audio quality: Auto Mode still uses Gemini TTS — quality stays identical to manual mode, just the voice/style varies.