minimax-tts

MiniMax TTS (WebSocket)

Trigger Conditions

• "Speech Synthesis", "TTS", "Text-to-Speech", "Voice Generation", "MiniMax TTS"

Overview

Perform speech synthesis via the MiniMax WebSocket API (wss://api.minimaxi.com/ws/v1/t2a_v2).

Supported Models

Model	Features
speech-2.8-hd	Accurately reproduces real tone details, comprehensively improving timbre similarity
speech-2.6-hd	Ultra-low latency, normalized upgrade, higher naturalness
speech-2.8-turbo	Accurately reproduces real tone details, faster and more affordable
speech-2.6-turbo	Speed version, suitable for voice chat and digital humans
speech-02-hd	Excellent prosody, stability, and replication similarity
speech-02-turbo	Enhanced capabilities for minority languages

Common Chinese Voices

Voice ID	Name
female-yujie	Mature Woman
female-shaonv	Young Girl
female-shaonv-jingpin	Young Girl-beta
female-tianmei	Sweet Female
qiaopi_mengmei	Playful Cute Girl
wumei_yujie	Charming Mature Woman
male-qn-qingse	Naive Youth
male-qn-jingying	Elite Youth
Cantonese_CuteGirl	Cute Girl (Cantonese)
Cantonese_GentleLady	Gentle Female Voice (Cantonese)

Full list: https://platform.minimaxi.com/docs/faq/system-voice-id

API Key

• Key source: openclaw.json → env.vars → MINIMAX_API_KEY
• Endpoint: wss://api.minimaxi.com/ws/v1/t2a_v2
• Authentication: Authorization: Bearer {api_key}

Usage

from tts_ws import synthesize

await synthesize(
    model="speech-2.8-hd",
    voice_id="female-yujie",
    text="你好世界",
    api_key="...",
    output_path="/path/to/output.mp3"
)

Notes

• Dependencies: websockets (pip3 install websockets)
• SSL: Currently skipping certificate verification (ssl.CERT_NONE)
• Clash TUN: api.minimaxi.com DIRECT rule is configured, WebSocket can connect normally
• Maximum 10,000 characters per request
• Output format: mp3 (32000Hz, 128kbps)