minimax/speech-2.6-turbo 📝🔢❓✓ → 🖼️

⭐ Official ▶️ 15.2K runs 📅 Oct 2025 ⚙️ Cog 0.16.8
low-latency multilingual-tts text-to-speech

About

Low‑latency MiniMax Speech 2.6 Turbo brings multilingual, emotional text-to-speech to Replicate with 300+ voices and real-time friendly pricing

Example Output

Output

Example output

Performance Metrics

3.30s Prediction Time
3.32s Total Time
All Input Parameters 
{
  "text": "Minimax just released Speech 2.6, It's really good, It builds on top of what existed before, The HD version is perfectly optimized for high-fidelity applications like voiceovers and audiobooks, And the Turbo variant is better for real-time applications with low latency.",
  "pitch": 0,
  "speed": 1,
  "volume": 1,
  "bitrate": 128000,
  "channel": "mono",
  "emotion": "auto",
  "voice_id": "Wise_Woman",
  "sample_rate": 32000,
  "audio_format": "mp3",
  "output_format": "hex",
  "language_boost": "None",
  "subtitle_enable": false,
  "english_normalization": false
}
Input Parameters
text (required) Type: string
Text to narrate (max 10,000 characters). Use markers like <#0.5#> to insert pauses in seconds.
pitch Type: integerDefault: 0Range: -12 - 12
Semitone offset applied to the voice (−12 to +12).
speed Type: numberDefault: 1Range: 0.5 - 2
Speech speed multiplier (0.5–2.0). Lower is slower, higher is faster.
volume Type: numberDefault: 1Range: 0 - 10
Relative loudness. 1.0 is default MiniMax gain. Range 0–10.
bitrate Default: 128000
MP3 bitrate in bits per second. Only used when audio_format is mp3.
channel Default: mono
mono for 1 channel (default), stereo for 2 channels.
emotion Default: auto
Desired delivery style. Use auto to let MiniMax choose, or pick a specific emotion.
voice_id Type: stringDefault: Wise_Woman
Voice to synthesize. Pick any MiniMax system voice or a voice_id returned by https://replicate.com/minimax/voice-cloning.
sample_rate Default: 32000
Audio sample rate in Hz.
audio_format Default: mp3
File format for the generated audio. Choose mp3 for general use, wav/flac for lossless, or pcm for raw bytes.
language_boost Default: None
Optional language hint. Choose Automatic to let MiniMax detect the language, or pick a specific locale.
subtitle_enable Type: booleanDefault: false
Return MiniMax subtitle metadata with sentence timestamps (non-streaming only).
english_normalization Type: booleanDefault: false
Improve number/date reading for English text (adds a small amount of latency).
Output Schema

Output

Type: string • Format: uri

Example Execution Logs 
Generating speech with model speech-2.6-turbo
Generated speech in 3.28sec
Each character is 1 token
Tokens: 270
Version Details
Version ID
24c0b2d2819faa5ce6eff09fc136c625e6e8c90e6f8a1cca75845f26fe9e1c4e
Version Created
November 7, 2025
Run on Replicate →