The automatic generation of video content model was unveiled at the re:Invent 2024 conference at the end of last year.New Reel, currently announced by AmazonCan generate an upgraded version with a duration of up to 2 minutes, and add more shooting angles.
The new version of Nova Reel 1.1 is built on the original Nova Reel technology. Users can use up to 4000 word prompts in Multishot Automated mode to generate 6 minutes of content from a 2-second video while maintaining a consistent shooting style.
In addition, the newly added Multishot Manual mode allows users to have greater control over the composition of generated videos. For example, by providing a single still image with a resolution of 1280 x 720 and generating a 6-second video from it, up to 512 lens frames can be generated through up to 20 groups of words, allowing users to choose the appropriate content.
Currently, the Nova Reel model is still only available through the AWS service platform, becoming an AWS exclusive application service project.
However, Amazon still hasn't disclosed the source of its own AI model training data, stating only that it is cross-trained through specific dedicated data and authorized data. However, Amazon stated in its service compensation terms that it will be responsible for any infringement allegations of content generated by its services and will step in to protect user rights.
