Model positioning
MiniMax Speech 2.5 is a speech generation model for multilingual text-to-speech and voice cloning, and a successor to Speech 02. It was officially released and announced for global availability on August 7, 2025. (minimax.io)
Capabilities and uses
- Multilingual speech synthesis: The official international announcement states that it supports more than 40 languages and offers a diverse voice library; performance in languages including Chinese and English has improved compared with the previous generation. (minimax.io)
- Voice cloning: MiniMax states that it can replicate accents, speaking styles, and emotions, and preserve a speaker's voice characteristics when generating speech across languages. Actual results may vary depending on the language and source material. (minimax.io)
- Synthesis controls: The official API documentation lists parameters including speech rate, volume, pitch, emotion, and audio format, and supports streaming output. Preview versions listed in the documentation include
speech-2.5-hd-previewandspeech-2.5-turbo-preview; these are specific version identifiers and do not change the null API ID in this record. (platform.minimaxi.com) - Use cases: Multilingual customer service, advertising and product voiceovers, localized content, short videos, and educational materials. (minimax.io)
Limitations
The request parameters in the official synchronous speech synthesis API documentation require input text to be fewer than 10,000 characters; limits may differ across APIs and versions. The documentation lists preview versions, and it cannot be used to confirm whether these versions are still available for use. Voice cloning should be performed with appropriate authorization from the voice rights holder. (platform.minimaxi.com)
Pricing and billing
As of November 18, 2026, MiniMax's official domestic and international pricing pages have been reviewed, but neither lists a usage-based unit price that can be clearly attributed to Speech 2.5, so pricing is unverified. The official domestic speech synthesis pricing table uses the number of input characters as the billing unit, and the API also returns the number of billable characters for the synthesis request; whether Speech 2.5 has a separate output audio charge is unverified. Do not treat prices for other models as prices for Speech 2.5. (platform.minimax.cn)