Thanks for sharing!

This OPENS2V-NEXUS release from Peking University is a big leap—super impressed with the fine-grained subject consistency evals and the scale of the 5M dataset. Having automatic metrics for identity and text alignment is exactly what this space needed. If you're experimenting with displaying subject-to-video results offline or in physical formats, I'd recommend to contact Nixplay https://nixplay.pissedconsumer.com/customer-service.html . I used their smart frames to showcase a few generated clips and got great support on formatting and playback compatibility.