Docs/Audio Models API

TTS

Audio Models API

Text-to-speech APIs powered by ElevenLabs and Minimax. Streaming TTS, async TTS, voice cloning and voice design.

Overview

The Audio API provides text-to-speech capabilities from multiple providers. ElevenLabs offers real-time streaming TTS, while Minimax (海螺) provides async TTS with voice cloning and voice design features.

ElevenLabs TTS

Real-time streaming TTS with natural voices. Supports 70+ languages.

Endpoints

POST/api/v1/tts/stream

Streaming text-to-speech conversion

GET/api/v1/voices

Get available voices

Available Models

Prices are per 1K characters

Model	Price	Description
eleven-tts-flash	¥0.20	Ultra-fast, lowest latency, 32 languages
eleven-tts-turbo	¥0.20	Fast, low latency, 32 languages
eleven-tts-multilingual	¥0.40	High quality, 29 languages
eleven-tts-v3	¥0.50	Most expressive, 70+ languages

TTS Request Parameters

modelrequiredstring

The model to use (see table above)

textrequiredstring

The text to convert to speech

voice_idstring

Voice ID. Default: 21m00Tcm4TlvDq8ikWAM (Rachel)

language_codestring

Language code (e.g., "en", "zh", "ja")

output_formatstring

Audio format. Default: mp3_44100_128

voice_settingsobject

Voice customization settings

Voice Settings

stabilitynumber

0-1. Higher = more consistent. Default: 0.5

similarity_boostnumber

0-1. Higher = more similar to original. Default: 0.75

stylenumber

0-1. Style exaggeration. Default: 0

use_speaker_boostboolean

Boost presence and clarity. Default: true

speednumber

0.7-1.2. Speech speed. Default: 1.0

ElevenLabs - Basic TTS

curl -X POST https://apimodels.app/api/v1/tts/stream \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "eleven-tts-flash",
    "text": "Hello, this is a test of text to speech.",
    "voice_id": "21m00Tcm4TlvDq8ikWAM"
  }' \
  --output audio.mp3

Response

The API returns a streaming audio response. The response headers include:

Content-Type: audio/mpeg
Transfer-Encoding: chunked
X-Credits-Charged: 0.02

ElevenLabs - With Voice Settings

curl -X POST https://apimodels.app/api/v1/tts/stream \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "eleven-tts-v3",
    "text": "This is expressive speech with custom settings.",
    "voice_id": "21m00Tcm4TlvDq8ikWAM",
    "language_code": "en",
    "output_format": "mp3_44100_192",
    "voice_settings": {
      "stability": 0.7,
      "similarity_boost": 0.8,
      "style": 0.3,
      "use_speaker_boost": true,
      "speed": 1.1
    }
  }' \
  --output audio.mp3

Popular Voices

Name	Voice ID	Description
Rachel	21m00Tcm4TlvDq8ikWAM	Conversational female
Adam	pNInz6obpgDQGcFmaJgB	Deep male
Antoni	ErXwobaYiN019PkySvjV	Friendly male
Bella	EXAVITQu4vr4xnSDxMaL	Warm female
Josh	TxGEqnHWrfWFTfGW9XjX	Deep young male
Sam	yoZ06aMxZJJ28mfd3POQ	Dynamic male

ElevenLabs - Get Voices

curl -X GET "https://apimodels.app/api/v1/voices?source=shared&page=0&page_size=30" \
  -H "Authorization: Bearer YOUR_API_KEY"

Voices Response

{
  "code": 200,
  "msg": "success",
  "data": {
    "voices": [
      {
        "voice_id": "21m00Tcm4TlvDq8ikWAM",
        "name": "Rachel",
        "category": "premade",
        "description": "Calm, friendly female voice",
        "labels": {
          "gender": "female",
          "age": "young",
          "accent": "american"
        },
        "preview_url": "https://..."
      }
    ],
    "has_more": true,
    "page": 0,
    "page_size": 30
  }
}

Output Formats

mp3_44100_128MP3, 44.1kHz, 128kbps (default)

mp3_44100_192MP3, 44.1kHz, 192kbps (higher quality)

pcm_16000PCM, 16kHz (raw audio)

pcm_22050PCM, 22.05kHz (raw audio)

pcm_44100PCM, 44.1kHz (raw audio)

Minimax TTS

Async text-to-speech by Minimax (海螺). Supports voice cloning, voice design, and pronunciation dictionaries.

Endpoints

POST/api/v1/minimax/tts/stream

Sync TTS - returns audio directly (best for short text)

POST/api/v1/minimax/tts

Async TTS - create task, poll for result (best for long text)

GET/api/v1/minimax/tts/status?task_id=xxx

Query async TTS task status

GET/api/v1/minimax/files/retrieve?file_id=xxx

Get file download URL

POST/api/v1/minimax/files

Upload audio file (voice clone)

POST/api/v1/minimax/voice/clone

Quick voice clone

POST/api/v1/minimax/voice/design

Voice design from description

GET/api/v1/minimax/voices

Get available voice list

Available Models

Prices are per 1K characters

Model	Price	Description
minimax-speech-02-turbo	¥3.20	Fast, cost-effective async TTS
minimax-speech-2.6-hd	¥5.60	High-definition, rich expressiveness

TTS Request Parameters

modelrequiredstring

minimax-speech-02-turbo or minimax-speech-2.6-hd

textrequiredstring

Text to convert to speech

voice_settingobject

Voice configuration (voice_id, speed, vol, pitch, etc.)

audio_settingobject

Audio output settings (sample_rate, bitrate, format, channel)

language_booststring

Boost specific language recognition (e.g., "zh", "en", "ja")

pronunciation_dictobject

Custom pronunciation dictionary

Minimax - Sync TTS (Stream)

Returns audio data directly. Best for short text (under ~5000 characters).

curl -X POST https://apimodels.app/api/v1/minimax/tts/stream \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "minimax-speech-02-turbo",
    "text": "Hello, this is Minimax text to speech.",
    "voice_setting": {
      "voice_id": "male-qn-qingse",
      "speed": 1.0,
      "vol": 1.0,
      "pitch": 0
    },
    "audio_setting": {
      "sample_rate": 32000,
      "bitrate": 128000,
      "format": "mp3"
    }
  }' \
  --output audio.mp3

Async TTS Workflow

POST /api/v1/minimax/tts to create task, get task_id

GET /api/v1/minimax/tts/status?task_id=xxx to poll status

When status is done, get file_id from response

GET /api/v1/minimax/files/retrieve?file_id=xxx to get download URL

Minimax - Async TTS

# Step 1: Create async TTS task
curl -X POST https://apimodels.app/api/v1/minimax/tts \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "minimax-speech-02-turbo",
    "text": "Hello, this is Minimax text to speech.",
    "voice_setting": {
      "voice_id": "male-qn-qingse",
      "speed": 1.0,
      "vol": 1.0,
      "pitch": 0
    },
    "audio_setting": {
      "sample_rate": 32000,
      "bitrate": 128000,
      "format": "mp3"
    }
  }'

TTS Response

{
  "task_id": "xxx-xxx-xxx",
  "base_resp": {
    "status_code": 0,
    "status_msg": "success"
  }
}

Status Response

{
  "task_id": "xxx-xxx-xxx",
  "status": 2,
  "file_id": "yyy-yyy-yyy",
  "base_resp": {
    "status_code": 0,
    "status_msg": "success"
  }
}

Voice Clone Workflow

POST /api/v1/minimax/files to upload audio sample (purpose: voice_clone)

POST /api/v1/minimax/voice/clone with file_id to create cloned voice

Use the cloned voice_id in TTS requests

First use of a cloned/designed voice for TTS incurs an additional ¥9.9 fee.

Minimax - Voice Clone

# Step 1: Upload audio sample
curl -X POST https://apimodels.app/api/v1/minimax/files \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F "purpose=voice_clone" \
  -F "file=@sample.mp3"

# Step 2: Clone voice (use file_id from step 1)
curl -X POST https://apimodels.app/api/v1/minimax/voice/clone \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "file_id": "FILE_ID_FROM_UPLOAD",
    "voice_id": "my-cloned-voice-001"
  }'

# Step 3: Use cloned voice in TTS
curl -X POST https://apimodels.app/api/v1/minimax/tts \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "minimax-speech-02-turbo",
    "text": "This uses my cloned voice.",
    "voice_setting": {
      "voice_id": "my-cloned-voice-001"
    }
  }'

Voice Design

Create custom voices from text descriptions. POST /api/v1/minimax/voice/design with a prompt describing the desired voice characteristics.

Minimax - Get Voice List

Returns the full Minimax system voice list (327 voices). Optionally filter by language with ?language=xxx.

curl -X GET "https://apimodels.app/api/v1/minimax/voices" \
  -H "Authorization: Bearer YOUR_API_KEY"

# Filter by language
curl -X GET "https://apimodels.app/api/v1/minimax/voices?language=English" \
  -H "Authorization: Bearer YOUR_API_KEY"

Response

{
  "voices": [
    {
      "voice_id": "male-qn-qingse",
      "name": "青涩青年音色",
      "language": "中文 (普通话)"
    },
    {
      "voice_id": "English_Graceful_Lady",
      "name": "Graceful Lady",
      "language": "English"
    }
  ],
  "total": 327
}

Supported language values: 中文 (普通话), 中文 (粤语), English, 日文, 韩文, 西班牙文, 葡萄牙文, 法文, 印尼文, 德文, 俄文, 意大利文, 阿拉伯文, 土耳其文 ...

Billing

ElevenLabs Billing

Credits based on character count:

Credits = (characters / 1000) * model_price * 10

Minimum charge: 10% of model price per request

Minimax Billing

Credits based on character count:

Credits = (characters / 1000) * credits_per_1K

Minimum charge: 10% of model price per request

Voice clone/design: additional ¥9.9 (99 credits) on first use of custom voice

Error Codes

400Invalid request parameters

401Invalid or missing API key

402Insufficient credits

404Model or voice not found

500Internal server error

View Models Get API Key