Text-to-Speech API 2026 - MiniMax Speech, Music & ASR (OpenAI-compatible)

🔊 Text-to-Speech & Audio models & pricing

Ten audio models: six MiniMax Speech TTS voices, two MiniMax Music song generators, and two speech-to-text (ASR) models - all on OpenAI-compatible audio endpoints.

Model	Price	Strength
minimax-speech-2.8-hd	~$0.054 / 1K chars	Latest flagship TTS, studio voices + emotion control	NEW
minimax-speech-2.8-turbo	~$0.031 / 1K chars	Fast low-latency TTS for realtime apps	NEW
minimax-speech-2.6-hd	~$0.054 / 1K chars	High-definition multilingual TTS	NEW
minimax-speech-2.6-turbo	~$0.031 / 1K chars	Cost-optimized TTS at scale	NEW
minimax-speech-02-hd	~$0.054 / 1K chars	Rich expressive timbres	NEW
minimax-speech-02-turbo	~$0.031 / 1K chars	Economical speech generation	NEW
minimax-music-v3.0	~$0.154 / song	Latest song generation, vocal or instrumental	NEW
minimax-music-v2.6	~$0.154 / song	Lyrics-to-song with configurable style	NEW
hy-asr-3.0-preview	~$0.003 / min	Hunyuan speech-to-text, timestamps + SRT	NEW
wand-asr-v1	~$0.007 / min	WAND multilingual ASR + language detection	NEW

Why run these models on NovAI?

Models

10 audio models

TTS from

~$0.031 / 1K chars

Music

~$0.154 / song

ASR from

~$0.003 / min

Output

URL or binary

Platform fee

$0

OpenAI-compatible /v1/audio/speech and /v1/audio/transcriptions - works with the OpenAI SDK
MiniMax Speech 2.8 HD adds emotion control and 32kHz studio-grade output
Return an audio URL (response_format=url) or raw mp3/wav/opus bytes
MiniMax Music generates full songs (vocal or instrumental) from lyrics + a style prompt
ASR returns text with timestamps plus an SRT subtitle URL

🚀 API usage

Text-to-speech — return an audio URL

curl https://aiapi-pro.com/v1/audio/speech \
  -H "Authorization: Bearer $NOVAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "minimax-speech-2.8-hd",
    "input": "Hello! Welcome to NovAI - one API key for every Chinese frontier model.",
    "voice": "male-qn-qingse",
    "response_format": "url"
  }'

Python (OpenAI SDK) — binary audio bytes

from openai import OpenAI
client = OpenAI(base_url="https://aiapi-pro.com/v1", api_key="YOUR_NOVAI_API_KEY")
resp = client.audio.speech.create(
    model="minimax-speech-2.8-hd",
    voice="male-qn-qingse",
    input="Hello! Welcome to NovAI.",
)
resp.stream_to_file("speech.mp3")

Speech-to-text (ASR)

curl https://aiapi-pro.com/v1/audio/transcriptions \
  -H "Authorization: Bearer $NOVAI_API_KEY" \
  -F model="hy-asr-3.0-preview" -F file="@audio.mp3"

TTS & Music use POST /v1/audio/speech (OpenAI-compatible). Set response_format to mp3/wav/opus for binary bytes (default) or url for a JSON body with audio_url (valid ~24h). Optional speed, volume, pitch, emotion. ASR uses POST /v1/audio/transcriptions and returns text with timestamps + an SRT subtitle URL. All audio is token-billed at $1.54 / 1M on real upstream usage.

Best use cases

Audiobook and podcast narration
IVR and voice-assistant responses
Video voice-over and dubbing
Accessibility / screen-reader voices
Music and jingle generation
Meeting and call transcription

Frequently asked questions

Is the TTS API OpenAI-compatible?

Yes. /v1/audio/speech follows the OpenAI audio shape, so client.audio.speech.create(model, voice, input) works with the OpenAI SDK - just point base_url to NovAI.

How is audio billed?

All audio (TTS, music, ASR) is token-billed at $1.54/1M on real upstream usage. Reference figures: HD TTS ~$0.054 per 1K chars, Turbo ~$0.031, music ~$0.154 per song, ASR ~$0.003-0.007 per minute.

What output formats are supported?

Set response_format to mp3, wav or opus for binary bytes (default), or url for a JSON body containing audio_url (valid ~24 hours).

Do you offer speech-to-text?

Yes - hy-asr-3.0-preview (~$0.003/min) and wand-asr-v1 (~$0.007/min, multilingual with language detection) via POST /v1/audio/transcriptions, returning text with timestamps and an SRT subtitle URL.