# 06 Encore

> Deepgram workshops › Build a voice agent › Mission 06
> Web version: https://workshops.deepgram.com/voice-agent/encore
> About 10 min (estimate). Check it: `make check 06` (offline, no key).

Push speed and expressivity until it does an impression.

- [Do the voice](https://workshops.deepgram.com/voice-agent/encore/do-the-voice.md)

## Do the voice

### Pick a bit

**Do:** Pick a bit, move **both** sliders, and tap **Start the agent**.

| The bit | Speed | Expressivity |
|---|---|---|
| Nature documentary narrator | 0.90 | 1 |
| Sports announcer in the final ten seconds | 1.15 | 2 |
| Calm pilot, mild turbulence | 0.95 | -1 |
| Bored teenager reading the fine print | 1.05 | -2 |
| Auctioneer selling your lunch | 1.15 | 2 |

_On the web page, this step has the Settings panel; in your own code the same settings are in the start file._

### Play it for someone

**Do:** Feed it "I am eating a sandwich," and play it for someone.

_On the web page, this step has the agent console: talk or type to the agent there, or run the same mission in your own code._

**Check:** You changed how it sounds.

_On the web page, this step shows the code panel with the matching code._

### Did it land?

**Do:** Did they laugh? That counts.

**Check:** The impression landed. (Optional.)

### Hard mode: two acts, one call

**Do:** **Hard mode.** Two new voices, one call, **Apply voice only**.

_On the web page, this step has the Settings panel; in your own code the same settings are in the start file._

**Check:** Hard mode: two acts, one call. (Optional.)

### Take it home

**What you tuned:** the voice layer. Same words, new delivery.

Code: [Python](https://github.com/deepgram-devs/voice-agent-workshop/blob/main/python/missions/06-encore/start.py) or [TypeScript](https://github.com/deepgram-devs/voice-agent-workshop/blob/main/typescript/missions/06-encore/start.ts).

**How it works**

Encore takes any agent from this workshop. The prompt starts as a nature documentary narrator; keep it or swap in a bit from the table above.

The first check reads two numbers from your session: `speed` and `expressivity`, next to the voice in the speak block. **Settings JSON** shows your values. The code shows how your app swaps voice and speed mid-call; Expressivity holds for the call, so change it and tap **Reconnect and apply**.

Polite smiles? Push Expressivity one further, or change the voice; the same prompt at the same speed sounds like a different performer.

**Hard mode:** pick a new **Voice** and **Speed**, tap **Apply voice only**, feed it a line; then once more with a third voice. The check counts swaps the agent answered after. Expressivity stays where the call started, so pick the act that suits it.

Speed and Expressivity are two lines in the Settings message. A ten-second edit to the voice layer is often what makes people play an agent for their friends.

The attendee project has three TODOs for the prompt, Speed, and Expressivity, and `make check 06` confirms them offline.

Headphones help: the difference between two Expressivity values is often in the breath and the pauses.

**In production**

- **Watch:** listener drop-off or "repeat that" requests after a voice change. Speed is the knob that moves them most.
- **Aim for:** Speed between 0.95 and 1.05 for anything people need to follow, such as numbers, addresses, and instructions, and save the extremes for performance.
- **Read:** the [Voice Agent TTS models](https://developers.deepgram.com/docs/voice-agent-tts-models) page for the voices and their options.
