Archive — US — techcrunch
Amazon says its AI video model can now generate minutes-long clips
Amazon has upgraded its AI video model, Nova Reel, allowing users to generate videos up to two minutes long. The new version, Nova Reel 1.1, features ‘multi-shot’ capabilities and a ‘Multishot Manual’ mode for better control. While it competes in a crowded generative video space, concerns over the ethical sourcing of training data linger, as Amazon hasn't disclosed this information.
Key points
- Amazon launched Nova Reel in December 2024 as its first generative video model.
- The updated Nova Reel 1.1 can create videos up to two minutes long and allows up to a 4,000-character prompt.
- New ‘Multishot Manual’ mode enables referencing an image to enhance video shot composition.
- Concerns exist regarding the ethical sourcing of training data for generative AI, particularly regarding copyright issues.
- Amazon has not disclosed the sources of its training data but offers legal protection for AWS customers against copyright claims.
Why it matters
- The rise of generative AI tools reflects significant trends in the tech industry towards automation and content generation.
- Similar generative models are being developed by OpenAI and Google, indicating competitive pressure and innovation in AI.
- Ethical considerations in AI training data sourcing are increasingly pertinent as companies grapple with copyright and intellectual property laws.
Amazon’s Nova Reel 1.1 signifies a significant advancement in generative video technology while raising important ethical questions about data sourcing and copyright protection in the evolving AI landscape.