SALESFORCE INTRODUCES TEXT-TO-VIDEO GENERATION

Summary from the TLDR AI newsletter

Salesforce has introduced xGen-VideoSyn-1, a text-to-video (T2V) model that generates realistic scenes from textual descriptions. The model uses a video variational autoencoder (VidVAE) to compress video data, reducing computational demands, and a Diffusion Transformer (DiT) for improved temporal consistency and generalization.