supertone-mcp

A composable MCP toolkit for the Supertone TTS API. Rather than a single "speak this text" command, it exposes Supertone's SDK as a set of building-block tools — synthesis, voice discovery, preview, duration/credit prediction, usage tracking, and full voice-cloning CRUD — that an LLM assembles to fulfill a request. Works in Claude Desktop, Cursor, or any MCP-compatible client.

Covers Korean, English, Japanese, and 31 languages total. Speed (0.5x–2.0x), pitch shift (-24 to +24 semitones), emotion styles, per-call output mode, streaming, and model selection.

Features

Synthesis

• text_to_speech — Convert text to audio. Per-call control of output_mode (files / resources / both), autoplay, streaming, model, plus include_phonemes / normalized_text. Long text is auto-chunked by the SDK.
• predict_duration — Estimate audio length (and credit cost) without synthesizing.

Voice discovery (preset)

• search_voice — Filter the catalog by language, gender, age, use_case, style, model, name, or description.
• get_voice — Full detail for one voice.
• preview_voice — Sample audio URLs for a voice (filterable by language/style/model).

Custom voice cloning

• clone_voice — Create a cloned voice from a local WAV/MP3 (≤3MB).
• search_custom_voice — List/filter cloned voices.
• get_custom_voice — Full detail for one cloned voice.
• edit_custom_voice — Update name and/or description.
• delete_custom_voice — Permanently delete (irreversible).

Audio assembly

• merge_audio_files — Concatenate two or more local audio files (mp3/wav) into one via a bundled ffmpeg. Supports plain concat, silence gaps between clips (gap_ms), or crossfade blending (crossfade_ms). Output format auto-detected (mixed → mp3) or forced via output_format. No system ffmpeg required.

Usage & credits

• get_credit_balance — Remaining credits.
• get_usage_history — Usage over a time window.
• get_voice_usage — Usage for a specific voice.

Breaking changes & migration (0.2.0)

0.2.0 moves behavior control out of environment variables and into per-call tool parameters — so the LLM decides per request, not the server config.

Before (env var)	After (per-call parameter)	Note
SUPERTONE_MCP_OUTPUT_MODE=files|resources|both	text_to_speech(output_mode=...)	Default still files
SUPERTONE_MCP_AUTOPLAY=true	text_to_speech(autoplay=...)	Default changed true → false (playback is now explicit)
(always streamed)	text_to_speech(streaming=...)	New, default false (one-shot). streaming=true requires model="sona_speech_1"

Other changes:

• Default model changed sona_speech_1 → sona_speech_2_flash.
• list_voices was removed (since the discovery release) and replaced by search_voice — call it with no arguments to reproduce the old "list everything" behavior.
• No more hard 300-character limit — longer text is auto-chunked by the SDK (credit/latency scale with length).

If you previously set SUPERTONE_MCP_OUTPUT_MODE or SUPERTONE_MCP_AUTOPLAY, remove them from your client config and pass output_mode / autoplay per call instead. (The server prints a one-time stderr notice if it sees the removed vars.)

Installation

# Using uvx (recommended)
uvx supertone-mcp

# Using pip
pip install supertone-mcp

Configuration

Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "supertone-tts": {
      "command": "uvx",
      "args": ["supertone-mcp"],
      "env": {
        "SUPERTONE_API_KEY": "your-api-key-here"
      }
    }
  }
}

Cursor

Add to your Cursor MCP settings (same JSON shape as above).

Environment Variables

Only authentication and stable defaults are configured via the environment — all behavior is controlled per call.

Variable	Required	Default	Description
SUPERTONE_API_KEY	Yes	—	Your Supertone API key
SUPERTONE_MCP_VOICE_ID	No	preset voice (Aiden, multilingual)	Default voice_id for text_to_speech / predict_duration (override per call)
SUPERTONE_OUTPUT_DIR	No	~/supertone-tts-output/	Directory where audio files are saved (used by output_mode=files/both)

Removed in 0.2.0: SUPERTONE_MCP_OUTPUT_MODE and SUPERTONE_MCP_AUTOPLAY — see Migration.

Output modes (text_to_speech output_mode)

Mode	Returns	Use when
files (default)	Plain text with the saved file path + metadata	You want the file on disk
resources	MCP AudioContent + TextContent (no file written)	The client renders audio inline (e.g., Claude.ai chat)
both	File on disk and AudioContent/TextContent	You want both — preview inline, keep the file

Usage Examples

The MCP client routes natural-language requests across these tools — the value of the toolkit is composition: the LLM chains several tools to satisfy one request.

Example 1 — Discover → preview → estimate cost → synthesize

"Find a calm Korean female voice, let me hear a sample, check the cost, then make this announcement as an mp3."

The LLM assembles:

search_voice(language="ko", gender="female", style="neutral")   # find candidates
  → preview_voice(voice_id)                                       # sample URLs to confirm the voice
  → predict_duration(text, voice_id) + get_credit_balance()       # gauge cost before spending
  → text_to_speech(text, voice_id, output_format="mp3",
                   output_mode="files")                           # synthesize

Example 2 — Clone my voice → use it right away

"Make a cloned voice from ~/recordings/sample.wav named MyVoice, then read this greeting with it and play it for me."

The LLM assembles:

clone_voice(name="MyVoice", audio_path="~/recordings/sample.wav")   # create the cloned voice
  → get_custom_voice(voice_id)                                       # confirm it was created
  → text_to_speech(text, voice_id=<cloned>, autoplay=true)           # synthesize, then play immediately

autoplay is a per-call parameter (default false), so playback happens only when explicitly requested.

Tool Parameters

text_to_speech

Parameter	Type	Required	Default	Description
text	string	Yes	—	Text to convert (long text is auto-chunked by the SDK)
voice_id	string	No	env or preset	Voice identifier (browse via search_voice)
language	string	No	ko	Language code — one of 31 (ko, en, ja, …)
output_format	string	No	mp3	mp3 or wav
model	string	No	sona_speech_2_flash	sona_speech_1, sona_speech_2, sona_speech_2_flash, sona_speech_2t, sona_speech_3t, supertonic_api_1, supertonic_api_3
speed	float	No	1.0	0.5–2.0
pitch_shift	int	No	0	-24 to +24 semitones
style	string	No	—	Emotion style (varies by voice)
output_mode	string	No	files	files, resources, or both (see Output modes)
autoplay	bool	No	false	Play the audio locally after synthesis (macOS afplay)
streaming	bool	No	false	Stream synthesis. Only supported by model="sona_speech_1"
include_phonemes	bool	No	false	Return phoneme timing data alongside the audio
normalized_text	string	No	—	Pre-normalized text (only used by sona_speech_2 / sona_speech_2_flash)

predict_duration

Same core parameter schema as text_to_speech (long text auto-chunked). Returns "Predicted duration: 2.34s (credit usage is proportional to duration).".

search_voice

All parameters optional. With no filters → full catalog. With any filter → first response line is Filters applied: ....

Parameter	Type	Description
language	string	e.g., ko, en, ja
gender	string	e.g., male, female
age	string	e.g., young_adult, child
use_case	string	e.g., narration, advertisement
style	string	e.g., neutral, happy
model	string	e.g., sona_speech_2_flash
name	string	partial match
description	string	partial match

get_voice / preview_voice

Tool	Required	Optional
get_voice	voice_id	—
preview_voice	voice_id	language, style, model (filter samples)

clone_voice

Parameter	Type	Required	Description
name	string	Yes	Display name (non-empty)
audio_path	string	Yes	Local WAV or MP3 path (≤3MB). Supports ~ expansion
description	string	No	Optional note

Custom voice CRUD

Tool	Required	Optional
search_custom_voice	—	name, description (partial match)
get_custom_voice	voice_id	—
edit_custom_voice	voice_id	name, description (at least one required)
delete_custom_voice	voice_id	— (IRREVERSIBLE)

Usage & credits

Tool	Required	Optional
get_credit_balance	—	—
get_usage_history	—	— (reports a recent default window)
get_voice_usage	voice_id	—

merge_audio_files

Parameter	Type	Required	Description
input_paths	string[]	Yes	Two or more local mp3/wav paths (~ expansion supported). A single file is returned as-is
gap_ms	int	No	Silence (ms) inserted at each junction. Default 0. Mutually exclusive with crossfade_ms
crossfade_ms	int	No	Crossfade blend (ms) at each junction. Default 0. Mutually exclusive with gap_ms
output_format	string	No	Force mp3 or wav. If omitted: all-same-ext → that ext; mixed → mp3

Development

# Clone and install
git clone https://github.com/supertone-inc/supertone-mcp.git
cd supertone-mcp
uv sync

# Run tests
uv run pytest -q

# Run with coverage
uv run pytest --cov=src --cov-report=term-missing

License

MIT