Users are now able to come up with prompts that are even up to 4,000 characters long and generate up to two-minute videos, which are composed of six-second shots. Nova Reel 1.1 also introduced a new model, which is called “Multishot Manual”.
With their latest creation, users who are generating videos in this mode are able to reference an image along with a prompt in order to offer more control over a video shot’s composition. Even more so, according to Fuentes, using a 1280 x 720 resolution image and a 512-maximum-character prompt for the Multishot Manual can even generate videos that contain up to 20 shots, as TechCrunch reported.
It is also worth mentioning that the Nova Reel is only available through AWS platforms and services, including Bedrock and Amazon’s AI dev suite, and customers need to request access to it. And, as with all generative artificial intelligence capabilities, there are some questionable queries on whether or not the Reel development used ethically sound methods.
Video-generating models are even trained on a vast number of example videos in order to “learn” the patterns in the videos without obtaining permission from the owners or creators, and, when these models “regurgitate” copyrighted stills, it exposes users of the models to IP lawsuits.
Amazon did not reveal its source for Reel Training data, nor has it provided an explicit way for creators to possibly offer the models with datasets with theri videos to opt out. However, the company also reported that they will protect any customer who is looking to be accused of violating copyright with media that was generated by their models.
.webp)
.webp)
