UT Austin releases code and model for video generation model StreamingT2V. Prior to this, the team published a paper on arXiv on March 22, describing an AI video generation technology called StreamingT2V, which is capable of generating highly consistent, length-scalable videos. According to the paper, StreamingT2V can generate long videos of 1,200 frames and 120 seconds in length, and theoretically even unlimited length.
GitHub Address:
https://github.com/Picsart-AI-Research/StreamingT2
Paper address:
https://arxiv.org/abs/2403.14773
