Amazon has upgraded its AI video model, Nova Reel, which can now generate videos up to two minutes in length.
Initially introduced in December 2024, Nova Reel marked Amazon’s entry into the generative video space. It now competes with similar technologies from OpenAI, Google, and other major players in the rapidly growing industry.
The latest version, Nova Reel 1.1, can create multi-shot videos while maintaining a consistent visual style throughout the clip. According to AWS developer Elizabeth Fuentes, users can input prompts of up to 4,000 characters to generate a video made up of six-second segments, with a total duration of up to two minutes.
This update also introduces a new feature called “Multishot Manual.” In this mode, users can provide a reference image (1280 x 720 resolution) along with a prompt of up to 512 characters to produce videos with up to 20 individual shots. This allows for greater control over the video’s content and composition.
Nova Reel is available exclusively through Amazon Web Services (AWS), including its AI development platform, Bedrock. Access to the tool requires special approval.
However, like many generative AI technologies, Nova Reel raises ethical concerns—especially around how the model was trained. Many AI models are trained on massive datasets that may include copyrighted videos without the creators’ consent. If the AI ends up reproducing such content, it can pose legal risks for users.
Amazon has not disclosed the source of Nova Reel’s training data, nor does it offer a clear opt-out method for creators whose content might have been used. Nevertheless, the company has stated it will legally protect AWS customers who face copyright infringement claims over content generated by its AI models, in line with its indemnification policy.