Provides offline text-to-speech by calling Ollama's /v1/audio/speech endpoint with the Phonon speech model. Enables speaking text aloud (queued/backgrounded playback), saving generated speech to wav/m4a/aiff files, listing available voices, and stopping playback, with Markdown/code/URL cleanup and sentence-boundary chunking of long text.
Click on "Deploy 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., "@ollama-tts-mcp Read the last paragraph aloud using the Zoey voice"
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.
Offline text-to-speech for Claude Code: a local alternative to the ElevenLabs MCP.
Speech is generated on your Mac by an open-source Ollama speech model
(Phonon, Apache 2.0, 112 MB) through Ollama's
/v1/audio/speech endpoint. After the one-time model pull, nothing touches the network.
Tool | What it does |
| Says text out loud. Runs in the background and queues behind earlier speech; |
| Saves speech to a file: |
| Lists the model's voices (Phonon has 38: Zoey, Hugh, Freya, Easton, Hadley, …). |
| Stops playback and clears the queue. |
Markdown, code blocks and URLs are cleaned up before speaking, and long text is split at sentence boundaries (the API takes at most 4,096 characters per request).
Related MCP server: Voice MCP
Apple Silicon Mac, Ollama 0.40.11 or newer (brew upgrade ollama)
ollama pull phonon (once)
Python 3.10+
If Ollama isn't running, the server starts ollama serve itself.
git clone https://github.com/HarsimarSingh23/ollama-tts-mcp.git
cd ollama-tts-mcp
ollama pull phonon
python3 -m venv .venv
.venv/bin/pip install "mcp>=1.9,<2"
claude mcp add --scope user ollama-tts -- "$PWD/.venv/bin/python" "$PWD/ollama_tts_mcp.py"Start a new Claude Code session afterwards; MCP servers are loaded when a session starts.
Set these with claude mcp add ... -e NAME=value:
Variable | Default | |
|
| Default voice |
|
| e.g. |
|
| Where |
|
| Ollama server |
|
| Start |
Phonon doesn't support speed or style instructions (Ollama returns an error for them).
No MP3 output; use m4a for small files.
Speech models currently run only on macOS in Ollama.
List the voices available in an Ollama speech model, and the speech models installed.
Args: model: Speech model tag to inspect. Defaults to the configured model.
Inputs: model. Returns: result
Say text out loud through this computer's speakers, fully offline.
Use it when the user asks you to speak, read something aloud, or announce a result.
Markdown, code blocks and URLs are tidied up before speaking. Playback runs in the
background and queues behind anything already playing; call stop_speaking to cut it off.
Args:
text: What to say. Any length; long text is split at sentence boundaries.
voice: Voice name, case-insensitive (e.g. "Zoey", "Hugh", "Freya"). Defaults to the
configured voice. See list_voices.
model: Ollama speech model tag, e.g. "phonon" or "phonon:multilingual". Defaults to the
configured model.
wait: If true, return only after the audio has finished playing.
Inputs: text (required), wait, model, voice. Returns: result
Stop the speech that is playing now and drop anything queued after it.
Inputs: no named parameters. Returns: result
Convert text to speech and save it as an audio file, fully offline. Does not play it.
Args:
text: The text to turn into speech. Any length.
voice: Voice name, case-insensitive (e.g. "Zoey", "Hugh"). See list_voices.
output_path: Where to save. Relative paths are resolved against the output directory
(default ~/Desktop). If omitted, a timestamped name is used.
format: "wav" (default), "m4a" (AAC, much smaller) or "aiff". If omitted, it is taken
from output_path's extension.
model: Ollama speech model tag. Defaults to the configured model.
strip_markdown: Remove code blocks, links and markdown markers before speaking.
Inputs: text (required), model, voice, format, +2 more. Returns: result
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 4 tools
Each tool has a clearly distinct purpose: speak plays audio, text_to_speech saves to a file without playing, list_voices enumerates voices/models, and stop_speaking halts playback. The play-vs-save distinction between speak and text_to_speech is explicitly called out, leaving little room for misselection.
list_voices and stop_speaking follow a clean verb_noun pattern, and speak is a simple verb, but text_to_speech is a noun-phrase outlier rather than an action verb like convert_text_to_speech. The set remains highly readable and predictable overall, with only that minor deviation.
Four tools is a well-scoped surface for an offline TTS server: one playback tool, one file-export tool, one discovery tool, and one interruption tool. Each earns its place with no redundancy.
The core lifecycle is covered: discover voices/models, speak aloud, export audio, and stop playback. Minor gaps exist around playback control (e.g. pause/resume, volume/speed adjustment, or checking playback status), but agents can work around these.
Synthesize short texts into WAV audio using offline formant voices in English, German and Spanish
Audio for your agent: transcribe, speak, translate, summarise, plus sound effects and music.
Remote text-to-speech MCP server with text_to_speech and list_voices tools. Free plan.
Text to speech and transcription for any AI model: MP3 voiceovers, audio and YouTube to text.
