Amazon Upgrades Nova Reel AI to Generate 2-Minute Multishot Videos with Enhanced Creative Control

Unveiled at the re:Invent 2024 conference late last year, Amazon’s Nova Reel—an AI-powered model for automatic video generation—has now been released in an upgraded version capable of producing clips up to two minutes long, while supporting an expanded array of camera angles.

Built upon the foundational technology of the original Nova Reel, version 1.1 introduces a Multishot Automated mode that leverages up to 4,000 token prompts. This enables users to extend video output from a mere 6 seconds to a full 2 minutes—while preserving a consistent visual style throughout.

The update also includes a new Multishot Manual mode, offering creators significantly greater control over composition. For instance, by inputting a single static image at 1280 x 720 resolution, users can generate a 6-second video sequence and use up to 512 tokens to produce up to 20 unique camera angles. From there, users can select the most compelling shots to craft their final sequence.

At present, Nova Reel remains exclusive to Amazon Web Services, cementing its place as a proprietary AWS offering.

While Amazon has not disclosed the exact training data behind its proprietary AI model, the company states that Nova Reel is trained using a combination of specialized and licensed datasets. Moreover, Amazon assures customers that it will assume responsibility for any claims of copyright infringement resulting from content generated via its service—pledging to defend and uphold user rights under its indemnification policy.