MiniMax Speech (international)

Overview

Best for: Expressive multilingual and Chinese-language voices, cloning and voice design where cost is secondary.

At a glance

$60/1M is speech-2.8-turbo; hd is $100/1M. 40 language_boost values. Rapid voice cloning $1.5 per voice. Limits are RPM by subscription, not concurrency. pcmu output at 8 kHz supported.

Features

• input streaming (t2a_v2_bidi)
• server-side sentence segmentation
• task_cancel for barge-in
• voice cloning
• voice design
• emotion control
• voice mixing

Pricing

Free tier: Not verified

Source: platform.minimax.io

Setup

1. Create an account on platform.minimax.io (international) and an API key.
2. Connect to wss://api.minimax.io/ws/v1/t2a_v2_bidi.
3. Wait for connected_success, send task_start (model, voice_setting, audio_setting), wait for task_started.
4. Send task_continue per text chunk, then task_finish; decode audio from data.audio.

Endpoint

wss://api.minimax.io/ws/v1/t2a_v2_bidi

Authentication

Authorization: Bearer <MINIMAX_API_KEY> (assumed from the standard T2A API; not shown on the bidi page excerpt)

# pip install websockets
import asyncio, json, os, websockets

URL = "wss://api.minimax.io/ws/v1/t2a_v2_bidi"
HDR = {"Authorization": f"Bearer {os.environ['MINIMAX_API_KEY']}"}

async def main():
    async with websockets.connect(URL, additional_headers=HDR) as ws:
        await ws.recv()  # connected_success
        await ws.send(json.dumps({"event": "task_start", "model": "speech-2.8-turbo",
            "voice_setting": {"voice_id": "YOUR_VOICE_ID", "speed": 1},
            "audio_setting": {"format": "pcm", "sample_rate": 24000}}))
        await ws.recv()  # task_started
        for t in ["Hello there. ", "This is streamed ", "from an LLM."]:
            await ws.send(json.dumps({"event": "task_continue", "text": t}))
        await ws.send(json.dumps({"event": "task_finish"}))
        with open("out_24k.pcm", "wb") as f:
            async for raw in ws:
                m = json.loads(raw)
                audio = (m.get("data") or {}).get("audio")
                if audio:
                    f.write(bytes.fromhex(audio))  # T2A returns hex-encoded audio
                if m.get("event") == "task_finished":
                    break

asyncio.run(main())

Written from the current docs. Check the vendor's SDK version before you ship.

Warnings

Plus 12 warnings that apply to all text-to-speech, streaming APIs. See category warnings.

Limits

• task_continue text under 10,000 characters
• ~120 s idle timeout; server sends no pings, client must
• One synthesis session per connection
• Subscription RPM: Starter 10 ... Business 800

Models and products

Docs and sources

Docs

• Bidirectional streaming
• Pay as you go pricing
• Audio subscription

Sources used

• platform.minimax.io/docs/guides/pricing-paygo.md
• platform.minimax.io/docs/guides/pricing
• platform.minimax.io/docs/api-reference/speech-t2a-websocket-bidi.md
• docs.livekit.io/reference/python/livekit/plugins/minimax/