This is a local Gemini-based Text-to-Speech (TTS) MCP server with automatic API key rotation, model fallback, and 30 multilingual voices. Here's what you can do:
Generate speech audio (generate_speech): Convert text into audio files (WAV, MP3, OGG, M4A, FLAC) with control over voice, speaking style, pitch factor, output path, and optional model override. Models automatically fall back between gemini-3.1-flash-tts-preview and gemini-2.5-flash-preview-tts.
Browse voices (list_voices): List all 30 multilingual voices with gender, tone, and description; optionally filter by gender and/or tone (e.g., "Soft", "Bright", "Firm", "Warm", "Upbeat").
Quick gender filter (list_voices_by_gender): Shortcut to list only male or female voices.
Manage API keys (reload_keys): Reload keys from ~/.gemini-tts-mcp/keys.json or environment variables without restarting.
Check key status (pool_status): See how many API keys are currently active in the rotation pool.
The server runs locally (keys never leave your machine) and offers automatic resilience with key rotation on quota/errors.
Provides text-to-speech generation using Gemini TTS models, with features like key rotation, model fallback, 30 multilingual voices, pitch control, and configurable speaking styles.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@gemini-tts-mcp say 'Hello, world!' in a friendly tone"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Multi-key Gemini TTS server (MCP) with automatic API key rotation and model fallback.
No baked-in voice, accent, or persona — all parameters are explicit.
🔄 Key rotation — pool of Gemini API keys; auto-rotates on quota/error
📉 Model fallback — tries gemini-3.1-flash-tts-preview → gemini-2.5-flash-preview-tts
🗣️ 30 voices — full catalog with gender, tone, and description
🔍 Filter by gender/tone — list_voices(male/female) or by tone name
🎚️ Pitch control — pitch_factor parameter (ffmpeg-based)
🌐 Multilingual — any voice speaks the language of your input text
🧩 MCP-native — register in any MCP host (Claude Desktop, Hermes Agent, etc.)
Related MCP server: gemini-webapi-mcp
⚠️ Not available on PyPI — install from source (see below).
pip install mcp google-genaiinstalls dependencies only, not this project.
Option A — Clone the repo (recommended):
git clone https://github.com/alarconcesar/gemini-tts-mcp.git
cd gemini-tts-mcp
pip install mcp google-genaiOption B — From anywhere (source must be in PYTHONPATH):
pip install mcp google-genai
# then clone & run from the repo directoryCreate ~/.gemini-tts-mcp/keys.json:
["AIzaSy...key1", "AIzaSy...key2"]Or set an env var: export GEMINI_API_KEY="AIzaSy...your_key"
cd gemini-tts-mcp
python -m gemini_tts_mcp.serverHermes Agent (~/.hermes/config.yaml):
mcp_servers:
gemini-tts:
command: "python"
args: ["-m", "gemini_tts_mcp.server"]Claude Desktop (claude_desktop_config.json):
{
"mcpServers": {
"gemini-tts": {
"command": "python",
"args": ["-m", "gemini_tts_mcp.server"]
}
}
}Tool | Description |
| Text → audio file (WAV, MP3, OGG, M4A, FLAC) |
| Browse all 30 voices (filter by gender/tone) |
| Shortcut: |
| Refresh API key pool from disk/env |
| Check how many keys are configured |
Param | Default | Description |
| (required) | Text to vocalize |
|
| Any of the 30 voices |
|
| Speaking style e.g. "softly", "cheerfully", "in a calm tone" |
|
| >1 = higher pitch, <1 = lower |
|
| Output format: |
|
| Override model (auto fallback if omitted) |
|
| Custom WAV path |
All 30 voices are multilingual — any voice speaks the language of your input text.
Female voices (16):
Voice | Tone | Description |
Achernar | Soft | Gentle, mellow tone |
Aoede | Breezy | Casual, relaxed delivery |
Autonoe | Bright | Clear, vibrant expression |
Callirrhoe | Easy-going | Laid-back, comfortable style |
Despina | Smooth | Refined, elegant tone |
Erinome | Clear | Crisp, distinct articulation |
Gacrux | Mature | Experienced, seasoned quality |
Kore | Firm | Assertive, confident delivery |
Laomedeia | Upbeat | Positive, energetic style |
Leda | Youthful | Young-sounding, fresh voice |
Pulcherrima | Forward | Direct, straightforward style |
Sterope | Forward | Bold, direct presentation |
Sulafat | Warm | Comforting, affectionate quality |
Vindemiatrix | Gentle | Soft, kind delivery |
Zephyr | Bright | High energy, clear articulation |
Male voices (15):
Voice | Tone | Description |
Achird | Friendly | Warm, approachable tone |
Algieba | Smooth | Polished, fluid delivery |
Algenib | Gravelly | Rough, textured quality |
Alnilam | Firm | Steady, resolute delivery |
Charon | Informative | Educational, explanatory style |
Enceladus | Breathy | Soft, airy quality |
Fenrir | Excitable | Energetic, animated expression |
Iapetus | Clear | Precise, well-articulated |
Orus | Firm | Strong, authoritative tone |
Puck | Upbeat | Cheerful, enthusiastic tone |
Rasalgethi | Informative | Educational, instructive |
Sadachbia | Lively | Energetic, spirited expression |
Sadaltager | Knowledgeable | Expert, well-informed tone |
Schedar | Even | Balanced, consistent tone |
Umbriel | Easy-going | Relaxed, conversational |
Zubenelgenubi | Casual | Informal, conversational |
Keys are loaded from ~/.gemini-tts-mcp/keys.json (a JSON array of strings), with env var fallback. The server rotates through the pool on each request, skipping failed keys per model. If all keys fail on one model, it falls back to the next model.
Use reload_keys() to refresh without restarting.
API keys never leave your machine — the MCP server runs locally
.gitignore excludes keys.json
keys.json.example provided as a template
MIT
Unclaimed servers have limited discoverability.
If you are the server author, to access and configure the admin panel.
MCP server for Google Veo AI video generation
MCP server exposing the AceDataCloud Fish Audio API (text-to-speech with voice conditioning)
MCP server for Hailuo (MiniMax) AI video generation
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/alarconcesar/gemini-tts-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server