Amazon has officially upgraded its Nova Reel AI model, enabling the generation of high-quality videos up to two minutes in length. This update, labeled Nova Reel 1.1, marks a significant expansion for the tool since its debut in December 2024, positioning Amazon as a major contender in the rapidly evolving generative video market against rivals like OpenAI and Google.
Advanced Multi-Shot Generation
The core improvement in Nova Reel 1.1 is its ability to produce “multi-shot” videos while maintaining visual consistency across different segments. Users can now input prompts of up to 4,000 characters to create a cohesive two-minute production, which the model structures using six-second shots.
New Manual Control Features
Amazon has introduced a “Multishot Manual” mode to provide creators with granular control over their output. By combining a reference image with a prompt, users can guide the composition of the video more effectively. When using a 1280 x 720-resolution image paired with a 512-character prompt, the model is capable of generating sequences containing up to 20 distinct shots.
Access and Ethical Considerations
Nova Reel remains exclusive to AWS platforms, including the Amazon Bedrock AI development suite. While developers must request access to the service, AWS currently approves these requests automatically.
Despite these technological strides, the model faces scrutiny regarding its training data. Generative video models learn by analyzing massive datasets of existing footage, raising concerns about the use of copyrighted material without creator consent. Amazon has not disclosed the specific sources used to train Nova Reel, nor has it implemented an opt-out mechanism for creators whose work may be included in the training sets.
To mitigate potential legal risks for its users, Amazon has confirmed it will extend its indemnification policy to protect AWS customers who face copyright claims stemming from media generated by its AI models.