MiniMax-Speech: Intrinsic Zero-Shot Text-to-Speech with a Learnable Speaker Encoder

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Bowen, Guo, Congchao, Yang, Geng, Yu, Hang, Zhang, Haozhe, Lei, Heidi, Mai, Jialong, Yan, Junjie, Yang, Kaiyue, Yang, Mingqi, Huang, Peikai, Jin, Ruiyang, Jiang, Sitan, Cheng, Weihua, Li, Yawei, Xiao, Yichen, Zhou, Yiying, Zhang, Yongmao, Lu, Yuan, He, Yucen
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916734887264256
author Zhang, Bowen
Guo, Congchao
Yang, Geng
Yu, Hang
Zhang, Haozhe
Lei, Heidi
Mai, Jialong
Yan, Junjie
Yang, Kaiyue
Yang, Mingqi
Huang, Peikai
Jin, Ruiyang
Jiang, Sitan
Cheng, Weihua
Li, Yawei
Xiao, Yichen
Zhou, Yiying
Zhang, Yongmao
Lu, Yuan
He, Yucen
author_facet Zhang, Bowen
Guo, Congchao
Yang, Geng
Yu, Hang
Zhang, Haozhe
Lei, Heidi
Mai, Jialong
Yan, Junjie
Yang, Kaiyue
Yang, Mingqi
Huang, Peikai
Jin, Ruiyang
Jiang, Sitan
Cheng, Weihua
Li, Yawei
Xiao, Yichen
Zhou, Yiying
Zhang, Yongmao
Lu, Yuan
He, Yucen
contents We introduce MiniMax-Speech, an autoregressive Transformer-based Text-to-Speech (TTS) model that generates high-quality speech. A key innovation is our learnable speaker encoder, which extracts timbre features from a reference audio without requiring its transcription. This enables MiniMax-Speech to produce highly expressive speech with timbre consistent with the reference in a zero-shot manner, while also supporting one-shot voice cloning with exceptionally high similarity to the reference voice. In addition, the overall quality of the synthesized audio is enhanced through the proposed Flow-VAE. Our model supports 32 languages and demonstrates excellent performance across multiple objective and subjective evaluations metrics. Notably, it achieves state-of-the-art (SOTA) results on objective voice cloning metrics (Word Error Rate and Speaker Similarity) and has secured the top position on the public TTS Arena leaderboard. Another key strength of MiniMax-Speech, granted by the robust and disentangled representations from the speaker encoder, is its extensibility without modifying the base model, enabling various applications such as: arbitrary voice emotion control via LoRA; text to voice (T2V) by synthesizing timbre features directly from text description; and professional voice cloning (PVC) by fine-tuning timbre features with additional data. We encourage readers to visit https://minimax-ai.github.io/tts_tech_report for more examples.
format Preprint
id arxiv_https___arxiv_org_abs_2505_07916
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle MiniMax-Speech: Intrinsic Zero-Shot Text-to-Speech with a Learnable Speaker Encoder
Zhang, Bowen
Guo, Congchao
Yang, Geng
Yu, Hang
Zhang, Haozhe
Lei, Heidi
Mai, Jialong
Yan, Junjie
Yang, Kaiyue
Yang, Mingqi
Huang, Peikai
Jin, Ruiyang
Jiang, Sitan
Cheng, Weihua
Li, Yawei
Xiao, Yichen
Zhou, Yiying
Zhang, Yongmao
Lu, Yuan
He, Yucen
Audio and Speech Processing
Sound
We introduce MiniMax-Speech, an autoregressive Transformer-based Text-to-Speech (TTS) model that generates high-quality speech. A key innovation is our learnable speaker encoder, which extracts timbre features from a reference audio without requiring its transcription. This enables MiniMax-Speech to produce highly expressive speech with timbre consistent with the reference in a zero-shot manner, while also supporting one-shot voice cloning with exceptionally high similarity to the reference voice. In addition, the overall quality of the synthesized audio is enhanced through the proposed Flow-VAE. Our model supports 32 languages and demonstrates excellent performance across multiple objective and subjective evaluations metrics. Notably, it achieves state-of-the-art (SOTA) results on objective voice cloning metrics (Word Error Rate and Speaker Similarity) and has secured the top position on the public TTS Arena leaderboard. Another key strength of MiniMax-Speech, granted by the robust and disentangled representations from the speaker encoder, is its extensibility without modifying the base model, enabling various applications such as: arbitrary voice emotion control via LoRA; text to voice (T2V) by synthesizing timbre features directly from text description; and professional voice cloning (PVC) by fine-tuning timbre features with additional data. We encourage readers to visit https://minimax-ai.github.io/tts_tech_report for more examples.
title MiniMax-Speech: Intrinsic Zero-Shot Text-to-Speech with a Learnable Speaker Encoder
topic Audio and Speech Processing
Sound
url https://arxiv.org/abs/2505.07916