transcript-tts
by ShevinuM
README.md
# transcript-tts
A minimal [MCP](https://modelcontextprotocol.io) server that converts text/transcripts into
MP3 audio using Microsoft Edge's free neural text-to-speech voices ([`edge-tts`](https://github.com/rany2/edge-tts)).
No API key, no account, no usage cap — it's an unofficial use of the same TTS service behind
Edge's "Read Aloud" feature.
## Tools
- **`text_to_speech(text, voice="en-US-AriaNeural", rate="+0%", volume="+0%", pitch="+0Hz", filename=None)`**
Synthesizes `text` to an MP3 file. Returns `{"audio_path": ..., "voice": ...}`.
Files are saved to `~/tts-output/` by default (override with `TRANSCRIPT_TTS_OUTPUT_DIR`).
- **`list_voices(locale_prefix=None)`**
Lists available voices, optionally filtered by locale (e.g. `"en-US"`, `"ja"`, `"en-GB"`).
## Setup
Requires [`uv`](https://docs.astral.sh/uv/).
```bash
git clone https://github.com/ShevinuM/transcript-tts.git
cd transcript-tts
uv sync
```
Register with Claude Code:
```bash
claude mcp add --scope user transcript-tts -- \
/path/to/transcript-tts/.venv/bin/python /path/to/transcript-tts/server.py
```
Or add it manually to your MCP client config (e.g. `claude_desktop_config.json`):
```json
{
"mcpServers": {
"transcript-tts": {
"command": "/path/to/transcript-tts/.venv/bin/python",
"args": ["/path/to/transcript-tts/server.py"]
}
}
}
```
## Environment variables
- `TRANSCRIPT_TTS_OUTPUT_DIR` — where MP3s are saved (default: `~/tts-output`)
- `TRANSCRIPT_TTS_VOICE` — default voice short name (default: `en-US-AriaNeural`)
## Notes
Requires `mcp[cli]>=2.0.0` — the MCP Python SDK renamed `FastMCP` to `MCPServer` in its 2.0
release, so this won't work against older `mcp` SDK versions without adjusting the import in
`server.py`.
TDQS
A4.2/5.0
Scored across 2 tools
Disambiguation5/5
The two tools have clearly distinct purposes: one lists available voices, the other converts text to speech. There is no overlap or ambiguity between them.
Naming Consistency5/5
Both tool names follow a consistent verb_noun pattern: 'list_voices' and 'text_to_speech'. The naming is predictable and uniform.
Tool Count3/5
With only two tools, the server feels minimal but appropriately scoped for a simple TTS purpose. It is on the lower end of the acceptable range as noted in the calibration guide.
Completeness4/5
The core workflow is covered: list voices to select one, then synthesize text. Minor gaps like streamed audio or additional output formats could be added, but the essential lifecycle is complete.
Maintenance
ActivityInactive
ResponsivenessNo recent community issues