Minimax Tts

MiniMax Text-to-Speech synthesis using the HTTP REST API. Generates high-quality audio from text in 40+ languages with ultra-realistic voices. Use when the u...

Security Scan

The skill's code and docs clearly require an API key and call the minimax.io REST API (expected for a TTS skill), but the registry metadata does not declare the required credential — an incoherence worth investigating before use.

Purpose & Capability

Name/description, SKILL.md, and scripts/tts.py all implement a TTS HTTP client to api.minimax.io — that capability is coherent. However the registry metadata lists no required environment variables or primary credential, while the SKILL.md and script require MINIMAX_API_KEY (or --api_key). The missing declared credential is an inconsistency.

Instruction Scope

SKILL.md instructs the agent to call the MiniMax REST endpoints (api.minimax.io / api-uw.minimax.io) and to run scripts/tts.py with provided flags. The instructions do not attempt to read unrelated files or credentials; they only reference MINIMAX_API_KEY (and an optional override).

Install Mechanism

No install spec is provided (instruction-only with an included script). No downloads or archive extracts are requested, so nothing is written to disk by an installer step beyond running the included script.

Credentials

The skill requires a bearer token (MINIMAX_API_KEY) to authenticate to the external TTS API — this is proportional for a remote TTS service. The concern is the skill registry metadata did not declare this required env var or a primary credential, which is a mismatch that could cause confusion or conceal required secrets.

Persistence & Privilege

always is false and there is no install step that requests persistent system-wide changes. The skill does not attempt to modify other skills or system configuration.

What to consider before installing

This skill appears to be a normal TTS client: it sends text to minimax.io and returns audio. However the package metadata failed to declare the required MINIMAX_API_KEY (the SKILL.md and script do require it). Before installing or using it: 1) Verify the service domain (api.minimax.io) and confirm the vendor is trustworthy. 2) Only provide a MINIMAX_API_KEY with least privilege and avoid sending sensitive text until you trust the service. 3) Ask the publisher or registry to correct the metadata to declare MINIMAX_API_KEY as a required credential. 4) You can review scripts/tts.py locally (it only POSTs JSON and writes returned audio bytes) and test with non-sensitive text first. If you are uncomfortable with sending data to an external API, do not install/use the skill.

MiniMax Text-to-Speech via HTTP REST API. Supports streaming and non-streaming, 40+ languages, 200+ voices.

API Details

• Endpoint: POST https://api.minimax.io/v1/t2a_v2
• Alt endpoint (lower latency): POST https://api-uw.minimax.io/v1/t2a_v2
• Auth: Bearer token via MINIMAX_API_KEY env var
• Content-Type: application/json

Quick Usage

uv run python scripts/tts.py --text "Hello world" --voice English_expressive_narrator --model speech-2.8-hd --output hello.mp3

Scripts

• scripts/tts.py — Core TTS script. Run with --help for full options.

Models

Model	Description
speech-2.8-hd	Ultra-realistic, supports sound tags
speech-2.8-turbo	Fast + natural flow
speech-2.6-hd	Low latency, enhanced naturalness
speech-2.6-turbo	Fast, affordable
speech-02-hd	Superior rhythm, high similarity
speech-02-turbo	Superior rhythm, multilingual

Output Formats

• mp3 (default), wav, pcm
• Sample rates: 32000 (default), 16000, 24000, 48000
• Bitrate: 128000 (default), 64000, 32000

Languages

40+ languages including: English, Chinese (Mandarin/Cantonese), Japanese, Korean, Spanish, French, German, Portuguese, Arabic, Russian, Hindi, Thai, Vietnamese, Turkish, Dutch, Polish, Italian, Indonesian, Malay, Persian, Swedish, Norwegian, Danish, Finnish, Hebrew, Romanian, Greek, Czech, Hungarian, Tamil, Afrikaans, and more.

Voices

Key English voices:

• English_expressive_narrator — Default expressive narrator
• English_radiant_girl — Radiant female
• English_magnetic_voiced_man — Magnetic male voice
• English_Aussie_Bloke — Australian male
• English_Whispering_girl — Whispering female
• English_PlayfulGirl — Playful female
• English_Comedian — Comedic voice
• English_AnimeCharacter — Female anime narrator

For full voice list (200+ voices across all languages), see references/voices.md.

Sound Tags (speech-2.8-hd only)

Use XML-like tags for breathing, pauses, expression:

• (sighs) — breathing sound
• (laughs) — laughter
• (coughs) — coughing
• [laughs] — laughing
• ... or (pause:500) — pause in ms
• <emphasis>important</emphasis> — emphasis
• <spell-out>A-P-I</spell-out> — spell out letters

Script Usage

uv run python scripts/tts.py --text "Your text here" [options]

Options:
  --text TEXT              Text to synthesize (required)
  --model MODEL           Model: speech-2.8-hd (default), speech-2.8-turbo, speech-2.6-hd, etc.
  --voice VOICE_ID        Voice ID (default: English_expressive_narrator)
  --speed SPEED           Speed 0.5-2.0 (default: 1.0)
  --pitch PITCH           Pitch -3 to 3 (default: 0)
  --vol VOLUME            Volume 0-10 (default: 1)
  --language_boost LANG   Language boost: auto (default), or specific lang e.g. en, zh
  --output_format FORMAT  hex (default) or raw (mp3/wav bytes returned directly)
  --format AUDIO_FORMAT   mp3 (default), wav, pcm
  --sample_rate RATE      32000 (default), 16000, 24000, 48000
  --bitrate BITRATE       128000 (default), 64000, 32000
  --stream                Enable streaming mode (returns chunks as they generate)
  --output FILE           Output file path (default: minimax_tts_output.mp3)
  --api_url URL           Override API URL
  --api_key KEY           Override API key (reads MINIMAX_API_KEY env if not set)

Streaming Mode

uv run python scripts/tts.py --text "Hello, streaming audio." --stream --output stream_output.mp3

Examples

# Basic
uv run python scripts/tts.py --text "The quick brown fox jumps over the lazy dog."

# Different voice
uv run python scripts/tts.py --text "Bonjour le monde" --voice French_Standard_Female --model speech-2.6-turbo

# Streaming
uv run python scripts/tts.py --text "This is streaming audio" --stream --output streaming.mp3

# With sound tags (expressive)
uv run python scripts/tts.py --text "Hello(sighs)... what a beautiful day(laughs)!" --voice English_expressive_narrator --model speech-2.8-hd