Overview
MiniMax Speech 2.8 is a text-to-speech model family focused on natural, expressive synthesized speech. MiniMax announced it on January 23, 2026; the supplied release identifier is Speech-2.8. The documented API variants are speech-2.8-hd and speech-2.8-turbo; these are variant identifiers, not a replacement for the unspecified model ID. (minimax.io)
Capabilities and uses
- Native sound tags can add effects such as laughter, breathing and sighing. The API also supports controls for voice, speed, volume, pitch and pronunciation. (minimax.io)
- MiniMax reports high-fidelity voice cloning from a 10-second sample and improved Mandarin-to-Japanese cross-lingual synthesis. Its current model documentation lists 40 supported languages and seven emotions. (minimax.io)
- Synchronous synthesis suits short conversational audio; asynchronous synthesis supports long-form text. Potential uses include narration, conversational interfaces and multilingual voice content. (platform.minimax.io)
Limitations
- The synchronous HTTP API requires text of fewer than 10,000 characters; MiniMax recommends streaming for texts over 3,000 characters. Asynchronous synthesis supports up to 1 million characters per request. (platform.minimax.io)
- Cross-language pronunciation and cloning fidelity are vendor-described improvements, not guarantees of identical or native-sounding output. Obtain appropriate permission before cloning a person's voice.
Official API pricing
Checked November 17, 2026. These are pay-as-you-go rates for text-to-audio synthesis, not token prices. The listed charge is based on text characters; the pricing tables do not list a separate generated-audio output rate. Synchronous and asynchronous rates are the same for each variant. Domestic and international offerings are separate and should not be treated as currency conversions. (platform.minimax.cn)
| Platform | Variant | Synthesis price |
|---|---|---|
| Mainland China | speech-2.8-hd |
CNY 3.50 per 10,000 input characters |
| Mainland China | speech-2.8-turbo |
CNY 2.00 per 10,000 input characters |
| International | speech-2.8-hd |
USD 100 per 1 million characters |
| International | speech-2.8-turbo |
USD 60 per 1 million characters |
For mainland billing, one Chinese character counts as two billable characters; English letters, punctuation, spaces and similar characters count as one each. Rapid voice cloning is separately priced at CNY 9.90 per voice domestically and USD 1.50 per voice internationally, charged when the cloned voice is first used for synthesis; preview text is billed at the selected model's rate. Resource packs may have different conditions and prices. (platform.minimax.cn)