wcy1122/

Long-TTS-Eval

This repository hosts the Long-TTS-Eval dataset, a benchmark released as part of the MGM-Omni project. It is designed for evaluating long-form and complex Text-to-Speech (TTS) capabilities, as well as speech and audio understanding in both English and Chinese.

Paper: MGM-Omni: Scaling Omni LLMs to Personalized Long-Horizon Speech GitHub: https://github.com/dvlab-research/MGM-Omni Demo: https://huggingface.co/spaces/wcy1122/MGM-Omni… See the full description on the dataset page: https://huggingface.co/datasets/wcy1122/Long-TTS-Eval.