MiniMax Speech 2.8
Speech 2.8 adds native sound tags for breaths, laughs and other spoken details. MiniMax announced the release on 23 January 2026.
Sound tags are part of the input
The announcement describes tags as modelled speech events rather than separate effects added after synthesis. They give the caller control over the rhythm and delivery of a spoken line.MiniMax Speech 2.8 announcement
Write for the listener
Practical guidance. Add only the sound cues that serve the script. Excessive breaths or laughter can distract from instructions. Compare HD and Turbo using the same voice, text and audio settings, then listen for missing words as well as tone.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | API, Provider product |
| Model identifiers | speech-2.8-hdspeech-2.8-turbo |
| Variants | HD and Turbo are hosted options within the same release.MiniMax current model specifications |
| Language coverage | The model catalogue lists 40 supported languages and seven emotions.MiniMax current model specifications |
| Interjection support | The HTTP API supports interjection tags specifically for Speech 2.8 HD and Turbo. A whisper emotion setting is not supported.MiniMax synchronous speech API |
Related models
- MiniMax Speech 02
MiniMax · MiniMax Speech
A hosted speech model with 24 languages, seven emotions and HD or Turbo serving options.
Launch date API - MiniMax ASR 1.0
MiniMax · MiniMax ASR
A transcription model for uploaded audio that can return speaker-labelled segments, timestamps or subtitle files.
Launch date not confirmedAPI - MiniMax Speech 2.6
MiniMax · MiniMax Speech
A speech model for responsive voice interaction and structured text, with Fluent LoRA to improve cloned-voice fluency.
Launch date not confirmedAPI