Fish Audio S2 原生支持多说话人特征自动处理

Title: Fish Audio S2 Natively Supports Automatic Multi-Speaker P...

URL Source: https://www.bestblogs.dev/status/2031685047099826257

Published Time: 2026-03-11 10:53:56

Markdown Content: 

原生多说话人支持 用户只需上传一段包含多个说话人的参考音频，模型通过 <|speaker:i|> token 自动处理每位说话人的特征，单次推理即可生成多人对话，无需再为每个说话人分别上传音频。

!Image 2: Tweet image

1 Replies

0 Retweets

0 Likes

1,378 Views 

One Sentence Summary

The S2 model supports handling multiple speaker characteristics in a single inference, eliminating the need for repeated audio uploads.

Summary

Following the S2 model announcement, this tweet details its 'Native Multi-Speaker Support.' Users only need to upload one reference audio containing multiple people; the model uses specific tokens like <|speaker:i|> to automatically identify and process each speaker's traits, enabling multi-person dialogue generation in a single inference and greatly simplifying complex voice production workflows.

AI Score

80

Influence Score 2

Published At Today

Language

Chinese

Tags

Fish Audio

S2

Multi-Speaker Support

Voice Inference

TTS