Aynursusuz/voice-design-bench-50-dnsmos

这是一个用于文本到语音任务的小型数据集，包含50个训练样本。每个样本包含文本和对应的三种不同音频格式（qwen_audio、echo_audio、omni_audio），以及它们的DNSMOS评分。音频采样率分别为24000Hz和44100Hz。数据集还提供了三种模型的DNSMOS评分统计信息。

This is a small dataset for text-to-speech tasks, containing 50 training samples. Each sample includes text and three corresponding audio formats (qwen_audio, echo_audio, omni_audio), along with their DNSMOS scores. The audio sampling rates are 24000Hz and 44100Hz. The dataset also provides DNSMOS score statistics for three models.