⌘K
Change language Switch ThemeSign In
Narrow Mode
Fish Audio S2 Open Source Details: 10 Million Hours of Data and Commercial Licensing ====================================================================================
Fish Audio S2 Open Source Details: 10 Million Hours of Data and Commercial Licensing ====================================================================================  ### 小互
@xiaohu
训练数据超过 1000 万小时音频，模型分为两个版本：S2（基础版）和 S2 Pro（旗舰版）
模型权重、微调代码、推理引擎全部开源，研究和非商用免费，但是商用需要联系授权。
详细介绍：xiaohu.ai/c/a066c4/fish-… 
Mar 11, 2026, 10:54 AM View on X
0 Replies
0 Retweets
1 Likes
1,302 Views  小互 @xiaohu
One Sentence Summary
Fish Audio reveals the training scale, versioning, and open-source licensing policy for the S2 model.
Summary
This tweet provides background data and open-source specifics for the S2 model. Trained on over 10 million hours of audio, it is available in two versions: S2 (Base) and S2 Pro (Flagship). The model weights, fine-tuning code, and inference engine are all open-sourced, following a 'free for research and non-commercial use, license required for commercial use' policy. A link for further exploration is provided for developers.
AI Score
82
Influence Score 1
Published At Today
Language
Chinese
Tags
Fish Audio
S2
Open Source License
Voice Model Training
S2 Pro HomeArticlesPodcastsVideosTweets
Fish Audio S2 Open Source Details: 10 Million Hours of Da... ===============