🤗 EMOVA-Models | 🤗 EMOVA-Datasets | 🤗 EMOVA-Demo
📄 Paper | 🌐 Project-Page | 💻 Github | 💻 EMOVA-Speech-Tokenizer-Github
EMOVA-Alignment-7M is a comprehensive dataset curated for omni-modal pre-training, including vision-language and speech-language alignment.
This dataset is created using open-sourced image-text pre-training datasets, OCR datasets, and 2,000 hours of ASR and TTS data.
This dataset is part of the EMOVA-Datasets… See the full description on the dataset page:
https://huggingface.co/datasets/Emova-ollm/emova-alignment-7m.
