Aynursusuz/tts-pretrain-refs-3k

这是一个用于文本转语音（TTS）预训练的参考数据集，包含3000个参考话语，每个话语对应一个独特的说话者（speaker_00000001至speaker_00003000）。音频采样率为44.1 kHz，以WAV格式存储在Parquet文件中。这些音频是由echo-tts合成的英语文本，基于Qwen3-TTS VoiceDesign基础说话者的潜在特征生成的。数据集的特征包括音频、文本和说话者ID，其中音频的采样率为44100 Hz。数据集分为一个训练集，包含3000个样本。配套数据集（clones）为Aynursusuz/tts-pretrain-clones-3m。

3000 reference utterances for TTS pretraining, one per speaker (speaker_00000001 through speaker_00003000). Sample rate: 44.1 kHz, WAV in Parquet. Generated by echo-tts synthesizing English text on speaker latents derived from Qwen3-TTS VoiceDesign base speakers. Features include audio, text, and speaker ID, with audio sampling rate of 44100 Hz. The dataset is split into a training set with 3000 examples. Companion dataset (clones): Aynursusuz/tts-pretrain-clones-3m.