Datasets:
| language: | |
| - en | |
| license: cc-by-4.0 | |
| pretty_name: E2 TTS Test Suite | |
| task_categories: | |
| - text-to-speech | |
| size_categories: | |
| - 1K<n<10K | |
| # E2 TTS Test Suite | |
| ## Summary | |
| This dataset converts the evaluation manifest from `microsoft/e2tts-test-suite` into Hugging Face Datasets format. | |
| Each example contains: | |
| - `text_prompt` | |
| - `transcription_of_audio_prompt` | |
| - `ground_truth_audio` | |
| - `audio_prompt_audio` | |
| - speaker metadata | |
| - `is_subjective_eval` | |
| ## Source | |
| Original repository: | |
| `https://github.com/microsoft/e2tts-test-suite` | |
| ## Notes | |
| - `ground_truth_audio` points to the reference utterance. | |
| - `audio_prompt_audio` points to the last 3 seconds prompt audio. | |
| - This dataset is intended for zero-shot TTS evaluation. | |
| ## License | |
| Please follow the original upstream data licenses. | |