# Voice, announcements, and TTS

## Configuration files

- **`adn-voice.yaml`** (optional, not committed) — merged into `config["VOICE"]`.
- Hot reload on file **mtime** (~15 s loop).

Template: `adn-voice.example.yaml`.

## Server voice identity (ID 5000)

All **server-originated** voice uses **RF source ID 5000** in the DMR stream so clients can tell infrastructure traffic from user radios:

- Scheduled **ANNOUNCEMENTS** and **TTS_ANNOUNCEMENTS** (destination = the **`TG`** field in each item).
- **On-demand** clips triggered by dialling **9991–9999** (playback uses destination **TG 9** on **TS2**; you still key **999x** to request the file).
- **Disconnected / reflector** voice lines (same: **TG 9** / **TS2**).

See [TG 9 — local service lane](special-numbers.md#tg-9-local-service-lane-prompts-and-bridge-plumbing) for why TG 9 is used and what must be enabled on the hotspot.
- **Voice ident** (destination **all-call** or **`OVERRIDE_IDENT_TG`**).

See [Special numbers — ID 5000](special-numbers.md#id-5000--server-voice-source-not-announcement-tg) for the full table.

## Features

| Feature | Description |
|---------|-------------|
| **Scheduled announcements** | AMBE phrases from `Audio/<lang>/ondemand/` on a schedule (hourly / interval). |
| **TTS announcements** | Text → gTTS → MP3 → ffmpeg → WAV → vocoder → AMBE (see pipeline below). |
| **Voice ident** | Periodic identification (optional); uses `VOICE_IDENT` in main server config and `OVERRIDE_IDENT_TG` / all-call. |
| **Recording** | When enabled, records AMBE to disk for traffic on **`RECORDING_TG`** / **`RECORDING_TIMESLOT`** (see template `adn-voice.example.yaml`). |
| **On-demand (9991–9999)** | See [Special numbers](special-numbers.md). |

## TTS pipeline (summary)

1. Source text in `.txt` under `Audio/<language>/ondemand/`.
2. **gTTS** produces **MP3**.
3. **ffmpeg** converts to **WAV**.
4. Vocoder command or **AMBEServer** produces **AMBE** for transmission.

**ffmpeg** must be installed on the host OS.

Configure **`TTS_VOCODER_CMD`** or **`TTS_AMBESERVER_HOST`** / **`TTS_AMBESERVER_PORT`** in `adn-voice.yaml` for encoding; see comments in `adn-voice.example.yaml`.

## Anti-collision (QSO) with announcements

When scheduling announcements, the server may **wait** if target slots are busy, **drop** targets if a live QSO appears mid-stream, and only mark **hourly** announcement state after a successful target list — this avoids clobbering live traffic.

## Broadcast queue

Parallel **broadcasts** on **different** TGs may run concurrently; **same-TG** broadcasts are **serialised** so one announcement completes before another on that TG.

## See also

- [Configuration](configuration.md) — voice file paths.
- [Echo](echo.md) — separate playback service.
- [Special numbers](special-numbers.md) — 5000, 999x, recording-related behaviour.