What is Amazon Nova Reel?
Amazon Nova Reel is a video-generation model provided by Amazon Web Services through Amazon Bedrock. It turns an English text prompt into video and can use a supplied JPEG or PNG image as visual guidance for the opening keyframe, subject, composition, or style.
The model is designed for producing visual sequences rather than text responses or general-purpose reasoning. Its current workflow uses the model identifier amazon.nova-reel-v1:1, which supports multi-shot generation and videos longer than the original six-second output. The earlier amazon.nova-reel-v1:0 model remains documented as a legacy version and has an announced end-of-life date of September 30, 2026.
Nova Reel fits within Amazon's Nova family of foundation models, but it serves a different purpose from text-oriented models. Its output is video, and its generation process is asynchronous: an application submits a job, waits for processing, and retrieves the completed files from Amazon S3.
What can Nova Reel create?
Nova Reel supports both text-to-video and image-guided video generation. A prompt can describe the subject, setting, action, lighting, camera movement, or cinematic direction. A reference image can provide a starting visual reference when a text description alone would not establish the desired composition or subject clearly enough.
- Short advertising and marketing concept videos.
- Product and brand visualizations.
- Social media video drafts.
- Storyboards and previsualization for film or commercial work.
- Short cinematic clips with specified camera movement.
- Image-guided sequences based on a supplied JPEG or PNG reference.
Nova Reel 1.1 also supports two different multi-shot approaches. Automated multi-shot generation can interpret one longer prompt and create a sequence of shots. Manual storyboard generation allows separate prompts and optional reference images for individual shots, giving the creator more control over the structure of the sequence.
Video specifications and limits
The model's output format is relatively fixed. It generates 1280x720 video at 24 frames per second, and duration is handled in six-second increments. Nova Reel 1.1 supports durations from six seconds up to 120 seconds, or two minutes.
| Specification | Verified detail |
|---|---|
| Provider | Amazon Web Services |
| Platform | Amazon Bedrock |
| Input modalities | English text and optional image |
| Output modality | Video |
| Resolution | 1280x720 |
| Frame rate | 24 fps |
| Duration | Six-second increments, up to 120 seconds with Nova Reel 1.1 |
| Standard prompt limit | Up to 512 characters for standard text-to-video generation |
| Automated multi-shot prompt limit | Up to 4,000 characters |
| Invocation | Asynchronous generation with Amazon S3 output |
The six-second increment is an important planning constraint. A request for a longer sequence is not equivalent to asking for an unrestricted continuous film: the generated result is organized around fixed shot intervals. Longer sequences may therefore require careful prompting, storyboard control, and review for continuity between shots.
How the Bedrock workflow works
Nova Reel does not use a conventional synchronous request that returns a completed video immediately. An application starts a generation job with Amazon Bedrock's asynchronous invocation operation, StartAsyncInvoke, and specifies an Amazon S3 destination. The service writes the generated video and related files to that location when processing is complete.
The asynchronous design makes Nova Reel suitable for queued creative pipelines, batch-style production, and applications that can notify users when a result is ready. It is less suitable for an interactive editor that expects an immediate preview after every prompt change.
A six-second generation typically takes about 90 seconds according to the supplied AWS documentation. A two-minute generation can take approximately 14 to 17 minutes. These timings are practical expectations rather than a guaranteed service-level performance figure, and actual completion time can vary.
Pricing and cost trade-offs
Nova Reel is not priced by input and output tokens. The supplied pricing information lists an output price of approximately $0.08 per generated video second, subject to AWS region, pricing tier, and current Amazon Bedrock pricing conditions.
At that listed rate, a six-second clip represents about $0.48 before any applicable variation in regional or account pricing. A 120-second generation represents about $9.60 on the same basis. These calculations describe the model's per-second generation charge only; they should not be treated as a complete estimate of every AWS cost associated with storage or the surrounding application.
The main cost-versus-speed trade-off is that Nova Reel offers a managed video-generation workflow without requiring users to operate their own generation infrastructure, but each iteration consumes paid generation time and may take minutes. It is therefore sensible to refine prompts and storyboard structure before submitting many long generations.
Strengths and limitations
Nova Reel's strongest practical feature is the combination of image guidance, camera-oriented prompting, and multi-shot generation in a managed Bedrock service. It can help teams move from a written creative brief or reference image to a visual draft without building a video-generation system from scratch. The manual storyboard mode is especially useful when shot-by-shot control matters more than generating an entire sequence from one description.
Its limitations are equally important. The model does not support audio input or audio output, so it is not a complete solution for synchronized speech, music, or sound design. It does not provide synchronous invocation, fine-tuning, or provisioned throughput in the supplied specifications. Output is restricted to 1280x720 at 24 fps, and the six-second shot structure limits the degree of direct control available to users seeking arbitrary durations or production formats.
Nova Reel should also be treated as a video-generation component rather than a nonlinear video editor. Generated footage may require review for visual consistency, subject continuity, safety, and suitability for a final production. The model's documentation does not provide a conventional textual knowledge-cutoff date, and no supplied benchmark establishes a general quality ranking against other video models.
Reasoning, coding, and tool support
Reasoning and coding are not meaningful primary capabilities of Nova Reel. The model is specialized for generating video from creative instructions rather than answering questions, writing software, or performing general-purpose analysis. No reasoning score, coding score, or context length is published in the supplied research.
Nova Reel does not expose general tool or function-calling support as a model capability. Amazon S3 and Bedrock asynchronous invocation are part of the surrounding delivery workflow, not evidence that the video model itself can independently call arbitrary tools or take actions.
When to choose Amazon Nova Reel
Choose Nova Reel when the goal is to create visual drafts, short cinematic clips, product concepts, advertising ideas, or storyboarded sequences from text and reference images, especially when the project already uses Amazon Bedrock or Amazon S3. Its current v1.1 workflow is the relevant choice when multi-shot generation or output longer than six seconds is required.
It is a better fit for asynchronous creative production than for real-time interaction. The model can be appropriate when a team values managed AWS integration and predictable output dimensions more than arbitrary resolution, instant previews, or built-in audio.
Another type of video tool may be more appropriate when the project requires synchronized dialogue or music, live generation, advanced editing, unrestricted frame rates and resolutions, or detailed control over every frame. A general-purpose language model is a better choice for scriptwriting, coding, research, or production planning; Nova Reel can then serve as the specialized video-generation stage after those materials are prepared.
Current lifecycle status
AWS documentation identifies amazon.nova-reel-v1:1 as the current workflow for multi-shot generation and longer videos. The original amazon.nova-reel-v1:0 model is legacy and is scheduled for end of life on September 30, 2026. New integrations should use the v1.1 model identifier where it is available and should account for the asynchronous S3-based workflow when designing their application.