A text-to-video model is a machine learning model which takes as input a natural language description and produces a video matching that description.[1]
Video prediction on making objects realistic in a stable background is performed by using recurrent neural network for a sequence to sequence model with a connector convolutional neural network encoding and decoding each frame pixel by pixel,[2] creating video using deep learning.[3]
There are different models including open source models. CogVideo presented their code in GitHub.[4] Meta Platforms uses text-to-video with makeavideo.studio.[5][6][7] Google used Imagen Video for converting text-to-video.[8][9][10][11][12]
Antonia Antonova presented another model.[13]
In March 2023, a landmark research paper by Alibaba research was published, applying many of the principles found in latent image diffusion models to video generation.[14][15] Services like Kaiber or Reemix have since adopted similar approaches to video generation in their respective products.
Matthias Niessner (TUM) and Lourdes Agapito (UCL) at AI company Synthesia work on developing 3D neural rendering techniques that synthesise realistic video. The goal is to improve existing text to video model by 2D and 3D neural representations of shape appearance and motion for controllable video synthesis of avatars that look and sound like real people.[16]
Although alternative approaches exist,[17] full latent diffusion models are currently regarded to be state of the art for video diffusion.