Custom TTS Service for Android

An Android Text-to-Speech (TTS) engine that connects to any OpenAI-compatible API backend (supporting the /v1/audio/speech endpoint).

This allows you to use custom, self-hosted, or cloud TTS voices — powered by AllTalk TTS, Piper, OpenAI, or any compatible service — as a standard system-wide TTS engine on your Android device.

---

Features

• Integrates as a standard Android TTS Engine (selectable in Android Settings → Text-to-Speech).
• Connects to any OpenAI-compatible /v1/audio/speech backend.
• API Key is optional — works without authentication for local servers.
• In-app Settings screen to configure:
  • Backend URL
  • API Key (optional, stored locally, input masked)
  • TTS Model
  • TTS Voice
  • Response Format (wav / mp3 / opus / pcm)
• Quick-fill presets for AllTalk (local) and OpenAI Cloud.
• Direct link to Android TTS engine selection from the main screen.
• Settings are persisted locally using Jetpack DataStore.

---

Setup

1. Install the app

Download the latest APK from Releases and install it, or build from source.

2. Select as TTS Engine

Open the app and tap "Open Android TTS Settings", then set Custom TTS as your preferred engine.

3. Configure the backend

Tap the ⚙️ icon to open Settings. Use a preset or fill in manually:

Field	AllTalk (local)	OpenAI Cloud
Backend URL	http://<your-pc-ip>:7851/v1/audio/speech	https://api.openai.com/v1/audio/speech
API Key	(leave empty)	Your OpenAI API Key
Model	piper	tts-1 or tts-1-hd
Voice	alloy	alloy, nova, echo, …
Format	wav	wav

Tap Save.

---

Local Setup with AllTalk TTS

AllTalk TTS runs on your PC/server and exposes an OpenAI-compatible API on port 7851.

1. Install and start AllTalk TTS on your machine.
2. Make sure your Android device and the server are on the same network.
3. Use the "AllTalk (Local)" preset in the app — update the IP address to match your server.

Note for WSL users: Set up a Windows port forward so the phone can reach WSL:

netsh interface portproxy add v4tov4 listenport=7851 listenaddress=0.0.0.0 connectport=7851 connectaddress=<WSL-IP>

---

Screenshots

---

Supported Audio Formats

Format	Status
wav	✅ Supported
pcm	✅ Supported
mp3	❌ Not yet implemented
opus	❌ Not yet implemented

---

Future Work / TODO

• MP3 / Opus decoding support.
• Implement onStop() to cancel ongoing network requests.
• More sophisticated language/voice mapping.