Amazon's Nova Reel 1.1 AI Model Generates Two-Minute Videos from Text Prompts

Amazon says its AI video model can now generate minutes-long clips | TechCrunch

Amazon has upgraded its AI video model, Nova Reel, with the ability to generate videos up to two minutes in length. Nova Reel, announced in December 2024, was Amazon's first foray into the generative video space. It competes with models from OpenAI, Google, and others in what's fast becoming a crowded market. The latest Nova Reel, Nova Reel 1.1, can generate "multi-shot" videos with "consistent style" across shots, explained AWS Developer Advocate Elizabeth Fuentes in a blog post. Users can provide a prompt up to 4,000 characters long to generate up to a two-minute video composed of six-second shots. Nova Reel 1.1 also introduces a new mode called "Multishot Manual." In this mode, the model can reference an image along with a prompt to offer more control over a video shot's composition. According to Fuentes, given a 1280×720-resolution image and 512-maximum-character prompt, Multishot Manual can generate videos containing up to 20 shots. Nova Reel is only available through AWS platforms and services including Bedrock, Amazon's AI development suite, and customers must request special access. As with most generative AI systems, there are questions as to whether Reel was developed using ethically sound methods. Video-generating models are trained on a vast number of examples of videos to "learn" the patterns in these videos to generate new clips. Some companies train models on copyrighted videos without obtaining permission from the owners or creators, and, when these models "regurgitate" copyrighted stills, it exposes users of the models to IP lawsuits. Amazon hasn't revealed the source of Reel's training data, nor has it provided an explicit way for creators possibly feeding the models' datasets with their videos to opt out. The company has, however, said that it'll protect any AWS customers accused of violating copyright with media generated by its models, in keeping with its indemnification policy.

Amazon's Latest Nova AI Model Can Generate 2-Minute-Long Videos

Users can either add a single prompt or multiple prompts for each shot Amazon introduced a new artificial intelligence (AI) video generation model that can create up to two minutes of video content with a single prompt. Announced on Monday, the Amazon Nova Reel 1.1 AI model is the successor to the Nova Reel model, which was introduced last year at its re:Invent event. While the latest video model comes with quality and latency improvements, the most significant upgrade is the total duration of video content it can generate in a single run. In a blog post, the tech giant announced the Nova Reel 1.1 AI model and detailed its capabilities. While the model can generate up to two minutes of videos, it will not be a continuous shot. The large language model (LLM) generates six-second-long shots, and a single video can have 20 such shots stitched together to reach the 120-second mark. The AI model is currently available to developers and general users via the Amazon Bedrock platform. The tech giant says users can either just add a single prompt and let the AI decide the shots for the final video or take control of the project and add specific prompts for each of the shots. The single prompt mode is dubbed Multishot Automated in Bedrock. Users can provide the model a prompt of up to 4,000 characters, and it will produce a multi-shot video adhering to the request. It only accepts text as input. The multi-prompt mode is dubbed Multishot Manual, which lets users add up to 512-character prompts for each shot. Users can also add an image as a reference for each shot. However, these images should have a resolution of 1280x720p. Amazon is pitching the AI model to those working in the advertising, marketing, media, and entertainment fields. The company says Nova Reel 1.1 comes with this customisability and lets users control elements such as camera motion, shot content, and more. Coming to pricing, the Amazon Nova Reel models (which includes both 1.0 and 1.1 versions) cost $0.08 (roughly Rs. 6.9) per second of video generated. This means one video of two minutes in length will cost $9.6 (roughly Rs. 830). This does not include reiterations, as that would be considered a separate project. Amazon does not offer inline edits in the videos. Notably, Amazon has not disclosed the source of the dataset fed to the Nova Reel AI model. However, the company does say that in case of a copyright violation lawsuit, it will protect all Amazon Web Services (AWS) clients as part of its indemnation policy.