Amazon is making waves in the generative AI landscape with a significant upgrade to its AI video model, Nova Reel. The new version 1.1 introduces a game-changing feature: the ability to generate realistic videos up to two minutes long. Available through AWS Bedrock, this update dramatically expands creative possibilities, positioning Amazon more competitively in the market and potentially disrupting the creative industry with high-quality, cost-effective AI video generation from text or images.

Conceptual illustration representing Amazon Nova Reel's AI using latent diffusion models to generate video content from prompts.

Nova Reel 1.1: Pushing the Boundaries of AI Video Generation

Launched in December 2024 at AWS re:Invent, Nova Reel marked Amazon’s first major entry into the generative video market. It now competes directly with advanced models from industry leaders like OpenAI (known for Sora’s cinematic capabilities), Google (with models like Imagen Video and Lumiere), and Runway (whose Gen-3 Alpha model served as an early benchmark). Amazon boldly claimed superiority from the start, stating that third-party evaluations showed Nova Reel outperformed Runway Gen-3 Alpha in both quality and consistency—a strategic move to establish credibility in this increasingly competitive space.

Nova Reel 1.1 addresses key limitations of its predecessor with transformative upgrades. The most impressive improvement is the ability to generate videos up to two minutes in length—a dramatic increase from the original six-second limit. As AWS Developer Advocate Elizabeth Fuentes explained, the model achieves this by intelligently connecting multiple related 6-second shots based on the user’s primary prompt, maintaining a “consistent style” throughout. This breakthrough enables more sophisticated storytelling, comprehensive product demonstrations, and rich narrative experiences.

Beyond longer videos, Nova Reel 1.1 gives creators substantially more control through expanded text prompts. Users can now provide instructions up to 4,000 characters long for standard two-minute videos—quadrupling the previous limit. This allows for remarkably detailed descriptions and specific directions, following AWS’s recommended prompting best practices. Creators gain precise control over scene elements, character actions, visual aesthetics, mood, and complex camera movements across multiple shots. Despite these advancements, Nova Reel 1.1 maintains the same high-quality output as its predecessor: 1280×720 pixel resolution at 24 frames per second, ensuring the “studio-quality” visuals that Amazon promises.

Introducing ‘Multishot Manual’ for Granular Control

Nova Reel 1.1 goes beyond just extending video length and prompt capabilities with its innovative ‘Multishot Manual’ feature. This specialized mode gives creators unprecedented control over video composition, particularly valuable for sequences requiring strong visual consistency or step-by-step scene building. It offers a more visually-driven approach compared to relying solely on text descriptions.

Here’s how the ‘Multishot Manual’ mode works, according to AWS:

Users start by providing a reference image (at 1280×720 resolution) and a text prompt (up to 512 characters in this mode) to generate the initial 6-second shot.
They can then add subsequent shots one at a time. According to Fuentes, this mode supports up to 20 individual shots, potentially reaching the two-minute maximum duration (20 shots × 6 seconds = 120 seconds).
For each additional shot, creators can either use the last frame of the previous shot as a visual reference or provide an entirely new image, along with a fresh prompt (limited to 512 characters per shot).

This methodical, image-guided approach gives creators direct influence over scene development on a shot-by-shot basis. It provides superior control over composition and visual continuity compared to the standard text-to-video method, enabling more precise realization of creative visions.

Conceptual graphic representing Amazon Nova Reel 1.1 AI video generation technology update

Strategic Positioning: Nova Reel in the AI Arms Race

This update represents more than technical improvements—it’s a strategic move within Amazon’s broader AI ambitions. The enhancement is part of the Amazon Nova family of foundation models, introduced alongside the original Nova Reel in December 2024. This comprehensive suite includes specialized models for text, images, and reasoning, all designed to integrate seamlessly with the AWS Bedrock platform. The approach demonstrates Amazon’s goal of becoming a complete AI solution provider, leveraging its powerful cloud infrastructure as a competitive advantage.

This positions Nova Reel 1.1 at the center of an intensely competitive AI video generation landscape, where innovation happens at breakneck speed. The stakes are enormous, as highlighted by market forecasts from Fortune Business Insights. They project the global AI video generator market will grow from approximately USD 614.8 million in 2024 to over USD 2.56 billion by 2032—a compound annual growth rate (CAGR) of 20.0%. This growth is fueled by the insatiable demand for video content across industries and AI’s potential to dramatically reduce production time and costs. By enabling longer videos and more detailed prompts with Nova Reel 1.1, Amazon enhances its appeal and strengthens its position within the AWS ecosystem, aiming to capture a significant portion of this rapidly expanding market.

Impact and Accessibility: What the Upgrade Means for Users

Nova Reel 1.1 represents more than technical advancement—it democratizes sophisticated video creation. For businesses and individual creators alike, these updates address critical limitations of earlier AI video tools and make professional-quality video production more accessible. The key impacts include:

Increased Accessibility: The ability to generate longer, complex videos faster and potentially more affordably lowers barriers to professional video production. This particularly benefits existing AWS customers who can seamlessly integrate Nova Reel into their current cloud environment. It reduces dependency on traditional resources like large production teams, expensive equipment, and lengthy post-production. At a reported cost of approximately $0.08 per second of video, Nova Reel offers a cost-effective alternative to competitors or conventional methods, especially considering the speed and reduced manual labor involved.
Enhanced Creative Control: The leap from six-second clips to two-minute videos enables more compelling narratives, detailed explanations, and comprehensive product demonstrations. This is further enhanced by the expanded 4,000-character prompt limit for standard generation and the precise, image-guided control of the new Multishot Manual mode. These features empower creators to realize more ambitious and nuanced creative visions.
Streamlined Workflows: Early adopters are already demonstrating Nova Reel’s practical applications across industries. Digital marketing agency Dentsu Digital uses it to accelerate ad campaign material creation. Musixmatch helps artists create distinctive music videos from lyrics. Stock media platform 123RF employs it to simplify visual media design. These examples highlight Nova Reel’s versatility in advertising, entertainment, media production, and potentially numerous other fields.

Despite these advantages, access to AWS Nova Reel remains restricted. The model is available exclusively through AWS platforms, primarily AWS Bedrock. Users need an active AWS account and must currently request specific permission to use the model. While this ensures seamless integration for existing AWS customers, it presents a barrier compared to platforms offering more open access. Bedrock utilizes an asynchronous API to efficiently manage the computationally intensive process of video generation in the background.

Behind the Pixels: Technology, Training Data, and Ethical Hurdles

Amazon Nova Reel is built on sophisticated technology, primarily utilizing latent diffusion models with transformer backbones. In simple terms, these are complex AI systems trained on vast datasets that learn intricate patterns. They interpret text or image inputs, convert them into an abstract internal representation (the ‘latent space’), and progressively construct the video frame by frame, transforming digital noise into coherent moving images that match the user’s specifications.

The effectiveness of these models heavily depends on the quality and volume of their training data. Amazon indicates that Nova Reel was trained on diverse sources, including licensed content, proprietary Amazon datasets, open-source materials, and significantly, large amounts of “publicly available information,” as detailed in their technical report. This last category, common to many large AI models, is at the center of ongoing industry debates. The lack of complete transparency about what constitutes “publicly available information” raises legitimate concerns about potentially using copyrighted material scraped from the internet without proper authorization or compensation.

Video-generating models, like their image-generating counterparts, learn by analyzing countless video samples to understand patterns and create new content. Some companies train their models on copyrighted videos without securing permission from rights holders. A significant concern is when these models effectively regurgitate training material rather than creating truly original content.

Amazon has not revealed the specific sources used in Reel’s training data, nor has it offered a clear mechanism for creators whose videos might be included in “publicly available” datasets to opt out. The company partially addresses intellectual property risks through its standard AWS service terms and general AI indemnification policy. Amazon has committed to defending AWS customers facing copyright infringement claims related to media generated by its models, providing some legal protection. However, this reactive approach doesn’t fully address the underlying ethical questions regarding data sources and creator rights, as transparent opt-out options remain scarce across the generative AI industry.

To proactively prevent misuse and promote responsible utilization, Amazon has integrated mandatory safety features directly into Nova Reel to filter potentially harmful or inappropriate content. While these measures represent important steps toward ethical AI video generation, the industry continues to grapple with balancing innovation and responsibility.