daremc86/serbian_common_voice
收藏资源简介:
这是一个塞尔维亚语单说话者语音数据集,采用Common Voice数据集格式风格,专门用于文本到语音(TTS)训练。数据集包含高质量的塞尔维亚语音录音,配有手动创建并以西里尔文字母编写的文本转录。语言为塞尔维亚语,使用西里尔文字母,格式与Common Voice兼容,适用于TTS任务。音频格式为WAV,采样率为22050 Hz,说话者为单一女性声音。数据集约8小时语音,经过清理和归一化处理,数字已从转录中移除,并已手动验证对齐。它兼容Coqui TTS、VITS、XTTS预处理管道和自定义PyTorch TTS管道,基于CC-BY-4.0许可证发布。
A Serbian single-speaker speech dataset prepared for Text-to-Speech (TTS) training using the Common Voice dataset formatting style. The dataset contains high-quality Serbian speech recordings paired with manually created text transcripts written in Cyrillic script. Language is Serbian, script is Cyrillic, format is Common Voice compatible, use case is Text-to-Speech, audio format is WAV, sampling rate is 22050 Hz, and speaker type is single-speaker female voice. Approximately 8 hours of speech, with cleaned and normalized transcripts, numbers removed, and manually verified alignments. Compatible with Coqui TTS, VITS, XTTS preprocessing pipelines, and custom PyTorch TTS pipelines, released under the CC-BY-4.0 license.



