AsianFin — Chinese AI startup MiniMax has rolled out a new update to its Speech-02 voice model, introducing a feature called “Voice Design,” according to a statement on Monday.

The new functionality allows users to describe their ideal voice characteristics using natural language, enabling precise, multi-dimensional control over voice outputs — including the ability to generate entirely novel voice tones that don’t exist in the real world.

Integrated seamlessly with the Speech-02 model, the Voice Design feature empowers users to achieve true “what you describe is what you get” performance in text-to-speech tasks, the company said.