MiniMax's New Voice Model Tops OpenAI & ElevenLabs!
MiniMax's Speech-02 TTS model beats OpenAI & ElevenLabs in voice AI! SOTA in cloning, multilingual support & cost efficiency.
"AI Disruption" Publication 6400 Subscriptions 20% Discount Offer Link.
The speed of progress in Chinese large-scale models has far exceeded people's expectations. At the beginning of the year, DeepSeek-R1 exploded in popularity, achieving performance that partially surpassed OpenAI’s o1 at an extremely low cost, to some extent dispelling the excessive "superstition" surrounding foreign large-scale models.
Now, in the field of voice AI, MiniMax, a heavyweight contender in the first tier of Chinese large-scale models, has dropped another "bombshell."
We see that its next-generation TTS voice model, Speech-02, has topped the international authoritative voice evaluation leaderboard, Artificial Analysis, decisively outperforming two industry giants, OpenAI and ElevenLabs!
It achieved state-of-the-art (SOTA) results in key voice cloning metrics, such as Word Error Rate (WER, lower is better) and Speaker Similarity (SIM, higher is better).
This achievement directly shocked foreign netizens, who exclaimed, “MiniMax will become a game-changer in the audio field.”
Renowned blogger AK also reposted about this new voice model:
Beyond its superior performance, Speech-02 is also highly cost-effective, with costs only one-fourth of ElevenLabs’ competing model (multilingual_v2).
The topping of Speech-02 once again demonstrates the technical strength and depth of Chinese large-scale models in surpassing top foreign competitors.
So, what magic does Speech-02 possess to achieve such remarkable results? With the release of the technical report this week, we delved into the technology behind the model.