[Suggestion] LiveKit Agents voice agent with Deepgram Flux TTS v2 expressivity and mid-stream control (Python)

[deepgram-robot]

What to build

A LiveKit Agents voice agent that demonstrates Deepgram's Flux TTS v2 expressivity features — dynamic speed adjustment, expressivity control, and barge-in interruption — through the recently merged livekit-plugins-deepgram TTS v2 plugin.

Why this matters

Developers building voice agents need nuanced TTS control: adjusting speaking pace during complex explanations, adding emotional expressivity for empathetic responses, and handling user interruptions gracefully. Deepgram's Flux TTS v2 supports mid-stream speed (0.85–1.15), expressivity (-2 to +2), and barge-in via send_interrupt() — unique capabilities in the LiveKit ecosystem. An example showcasing these features helps developers leverage the full Flux TTS v2 surface area within LiveKit Agents rather than relying on static TTS configuration.

Suggested scope

• Python using livekit-agents with livekit-plugins-deepgram (Flux TTS v2)
• Demonstrates: per-turn expressivity adjustment, speed variation based on content complexity, barge-in handling with playback_offset
• Uses Deepgram STT (Nova-3) + Deepgram TTS (Flux v2) + an LLM provider
• Includes a LiveKit Agents web frontend for testing (room join UI)
• Shows send_configure() for mid-stream parameter changes and ConfigureSuccess/ConfigureFailure acknowledgment handling

Acceptance criteria

• Runnable with minimal setup (clone, add API key, run)
• README explains the pattern clearly
• Uses current SDK version
• Demonstrates at least 3 Flux TTS v2 features (speed control, expressivity, barge-in)
• Shows mid-stream TTS parameter reconfiguration during a live conversation
• Frontend displays TTS parameter changes in real-time

---

Raised by the DX intelligence system.