OpenS2S_ Datasets
收藏资源简介:
OpenS2S数据集是一个完全开源的、透明且端到端的LSLM,旨在实现富有同情心的语音交互。该数据集基于BLSP-Emo模型,采用流式交织解码架构以实现低延迟语音生成。OpenS2S还包含一个自动化的数据构建管道,能够以低成本合成多样化的、高质量的同情心语音对话。通过利用大型语言模型生成同情心内容和可控的文本到语音系统引入说话者和情感变化,构建了一个具有丰富副语言多样性和最小人工监督的可扩展训练语料库。该数据集为研究社区提供了丰富的资源,包括数据集、模型权重、预训练和微调代码,以促进合作研究和推动同情心语音系统的创新。
The OpenS2S dataset is a fully open-source, transparent, end-to-end LSLM designed to enable compassionate speech interactions. Built on the BLSP-Emo model, this dataset adopts a streaming interleaved decoding architecture to achieve low-latency speech generation. OpenS2S additionally features an automated data construction pipeline that can synthesize diverse, high-quality compassionate speech dialogues at low cost. By leveraging large language models to generate compassionate content and controllable text-to-speech systems to introduce speaker and emotional variations, it constructs a scalable training corpus with rich paralinguistic diversity and minimal human supervision. This dataset provides abundant resources for the research community, including the dataset itself, model weights, pre-training and fine-tuning codes, to facilitate collaborative research and drive innovation in compassionate speech systems.




