instinct-org/default_voices_chunked_speech_restorised_tts_train_clone_pairs

default_voices_chunked_speech_restorised_tts_train_clone_pairs This is a gated Uzbek TTS training clone-pair dataset. It contains tokenized speaker-reference and target pairs for text-to-speech voice adaptation workflows. Language Primary language: uz (Uzbek) Contents audios/shard-*.tar: tokenized audio shards txts/shard-*.jsonl: per-example metadata and text fields data.lst: repository-relative shard manifest tokenized_dataset_summary.json:… See the full description on the dataset page: https://huggingface.co/datasets/instinct-org/default_voices_chunked_speech_restorised_tts_train_clone_pairs.