✨T2AV-Compass: Towards Unified Evaluation for Text-to-Audio-Video Generation

📝 Summary:
T2AV-Compass introduces a unified benchmark for text-to-audio-video generation evaluation. It features 500 diverse prompts and a dual-level framework. Evaluations reveal current T2AV models struggle significantly with realism and cross-modal consistency.

🔹 Publication Date: Published on Dec 24

🔹 Paper Links:
• arXiv Page: https://arxiv.org/abs/2512.21094
• PDF: https://arxiv.org/pdf/2512.21094
• Project Page: https://nju-link.github.io/T2AV-Compass/
• Github: https://github.com/NJU-LINK/T2AV-Compass/

==================================

For more data science resources:
✓ https://t.me/DataScienceT

#TextToAudioVideo #MultimodalAI #AIEvaluation #GenerativeAI #AIResearch