Dataset · ASLP-lab
Benchmark for contextual and multi-speaker speech generation
ASLP-lab/DRAMA-Eval
A bilingual audio and text dataset for evaluating speech synthesis models in multi-turn dialogues with multiple speakers.
What's inside
The repository contains evaluation materials for the DRAMA-Eval benchmark: ground-truth audio recordings, text scenarios, and audio prompts in wav format. The dataset size ranges from 1 to 10 thousand records in English and Chinese. The data is designed to test how naturally text-to-speech (TTS) models maintain context in dialogues.
What you can do with it
- Evaluating the quality of conversational TTS models and voice cloning systems
- Testing speech models' ability to convey intonation and context in multi-speaker scenes
- Tasks
- Text → speech
- Languages
- English, Chinese
- Size
- 1K–10K rows
- Downloads this month
- 3.9K
- Likes
- 16
How to load it
from datasets import load_dataset
ds = load_dataset("ASLP-lab/DRAMA-Eval")