| license: apache-2.0 | |
| task_categories: | |
| - text-to-speech | |
| language: | |
| - en | |
| - ja | |
| size_categories: | |
| - 1M<n<10M | |
| configs: | |
| - config_name: default | |
| data_files: | |
| - split: train | |
| path: | |
| - data/train-*.parquet | |
| - data2/train-*.parquet | |
| - data3/train-*.parquet | |
| dataset_info: | |
| config_name: default | |
| features: | |
| - name: audio | |
| dtype: audio | |
| - name: text | |
| dtype: string | |
| - name: speaker_id | |
| dtype: string | |
| splits: | |
| - name: train | |
| num_examples: 5957779 | |
| # ja-en-syn | |
| Merged synthetic TTS clone dataset: 5957779 utterances total. | |
| - **English** (`data/`, `data2/`) - 2967779 rows from `tts-pretrain-clones-3m`. | |
| - **Japanese** (`data3/`) - 2990000 rows from `irodori-clones-3m`, with style | |
| emoji in the `text` field converted to `<tag>`s (`<whisper>`, `<breath>`, | |
| `<slow>`, ...) and every other emoji removed. | |
| Columns: `audio` (Audio), `text` (string), `speaker_id` (string). | |