MINT-Bench: A Comprehensive Multilingual Benchmark for Instruction-Following Text-to-Speech
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866917422193180672 |
|---|---|
| author | Chen, Huakang Hu, Jingbin Xue, Liumeng Zhan, Qirui Li, Wenhao Ma, Guobin Xie, Hanke Guo, Dake Ma, Linhan Jiang, Yuepeng Wu, Bengu Xie, Pengyuan Xie, Chuan Zhang, Qiang Xie, Lei |
| author_facet | Chen, Huakang Hu, Jingbin Xue, Liumeng Zhan, Qirui Li, Wenhao Ma, Guobin Xie, Hanke Guo, Dake Ma, Linhan Jiang, Yuepeng Wu, Bengu Xie, Pengyuan Xie, Chuan Zhang, Qiang Xie, Lei |
| contents | Instruction-following text-to-speech (TTS) has emerged as an important capability for controllable and expressive speech generation, yet its evaluation remains underdeveloped due to limited benchmark coverage, weak diagnostic granularity, and insufficient multilingual support. We present \textbf{MINT-Bench}, a comprehensive multilingual benchmark for instruction-following TTS. MINT-Bench is built upon a hierarchical multi-axis taxonomy, a scalable multi-stage data construction pipeline, and a hierarchical hybrid evaluation protocol that jointly assesses content consistency, instruction following, and perceptual quality. Experiments across ten languages show that current systems remain far from solved: frontier commercial systems lead overall, while leading open-source models become highly competitive and can even outperform commercial counterparts in localized settings such as Chinese. The benchmark further reveals that harder compositional and paralinguistic controls remain major bottlenecks for current systems. We release MINT-Bench together with the data construction and evaluation toolkit to support future research on controllable, multilingual, and diagnostically grounded TTS evaluation. The leaderboard and demo are available at https://longwaytog0.github.io/MINT-Bench/ |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_17958 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | MINT-Bench: A Comprehensive Multilingual Benchmark for Instruction-Following Text-to-Speech Chen, Huakang Hu, Jingbin Xue, Liumeng Zhan, Qirui Li, Wenhao Ma, Guobin Xie, Hanke Guo, Dake Ma, Linhan Jiang, Yuepeng Wu, Bengu Xie, Pengyuan Xie, Chuan Zhang, Qiang Xie, Lei Audio and Speech Processing Sound Instruction-following text-to-speech (TTS) has emerged as an important capability for controllable and expressive speech generation, yet its evaluation remains underdeveloped due to limited benchmark coverage, weak diagnostic granularity, and insufficient multilingual support. We present \textbf{MINT-Bench}, a comprehensive multilingual benchmark for instruction-following TTS. MINT-Bench is built upon a hierarchical multi-axis taxonomy, a scalable multi-stage data construction pipeline, and a hierarchical hybrid evaluation protocol that jointly assesses content consistency, instruction following, and perceptual quality. Experiments across ten languages show that current systems remain far from solved: frontier commercial systems lead overall, while leading open-source models become highly competitive and can even outperform commercial counterparts in localized settings such as Chinese. The benchmark further reveals that harder compositional and paralinguistic controls remain major bottlenecks for current systems. We release MINT-Bench together with the data construction and evaluation toolkit to support future research on controllable, multilingual, and diagnostically grounded TTS evaluation. The leaderboard and demo are available at https://longwaytog0.github.io/MINT-Bench/ |
| title | MINT-Bench: A Comprehensive Multilingual Benchmark for Instruction-Following Text-to-Speech |
| topic | Audio and Speech Processing Sound |
| url | https://arxiv.org/abs/2604.17958 |