JSON Layered Video Dataset for Scene Structure & Compositing
Most video datasets expose only the final rendered pixels. This collection also preserves structured information describing how the visual scene was assembled.
Access more than 700,000 layered video compositions representing approximately 2,917 hours of structured visual content. Each composition typically combines 2–4 visual elements such as base footage, HUD graphics, overlays, atmospheric effects, flares, particles, bokeh or graphs. Associated JSON describes the layers, composition structure, asset relationships, positioning and rendering information, connecting the visible result with the structure used to construct it.
This combination makes the dataset suitable for models that need to learn scene decomposition, visual compositing, element placement and relationships between component assets rather than treating every rendered frame as a single flattened image.
Dataset Preview
Representative rendered compositions showing scenes assembled from multiple visual elements including base footage, interface graphics, overlays, particles, atmospheric effects, flares, bokeh and graphs.
Each preview represents a final visual composition whose underlying layer structure and asset relationships are described in the associated composition data, allowing the rendered result to be studied together with information about how the scene was assembled.
Key Highlights
- • 700,000+ structured layered video compositions
- • Typical compositions contain 2–4 independently represented visual layers
- • JSON preserves composition structure instead of exposing only flattened output
- • Layer relationships and positioning connect component elements to the assembled scene
- • Visual elements include base video, graphics, overlays, particles, flares, bokeh and atmospheric effects
- • Rendered previews provide a visual reference for the corresponding structured composition records
Metadata Fields
Technical Specifications
Dataset Type
Structured Layered Video + JSON Composition Dataset
Library Size
700,000+ compositions
Total Duration
Approximately 2,917 hours
Average Composition Duration
Approximately 15 seconds
Typical Layer Count
2–4 visual layers per composition
Metadata Format
Structured JSON
Composition Structure
Structured relationship between multiple visual elements and the assembled rendered scene
Layer Types
Base video, HUD graphics, overlays, atmospheric effects, flares, particles, bokeh, graphs, and other composited assets
Preview Assets
Rendered composition previews
Asset Relationships
Preserved within structured composition metadata
Positioning Information
Available within composition JSON
Rendering Information
Available within composition JSON
Audio
None
Licensing, Documentation, and Delivery
Wavebreak Media can provide applicable licensing terms, dataset documentation, and structured JSON metadata information for the layered video composition content, with dataset delivery organized around the intended model-development, research, retrieval, or visual-generation workflow. See our Dataset Licensing & Compliance and Dataset Delivery & Security pages for details on rights documentation, packaging, and secure transfer.
Need the Full Dataset?
Request a representative sample containing rendered composition previews and corresponding JSON records to evaluate layer structure, positioning, asset relationships and rendering information before selecting a larger dataset volume.
Related Datasets
Adobe After Effects Source Projects Dataset
License 500+ native Adobe After Effects project files with editable compositions, layers, timelines, keyframes, previews, and project metadata for creative AI.
View Dataset EDITABLE TEMPLATE DATASETAfter Effects Templates Dataset
Access up to 300 editable After Effects templates with structured production metadata for creative AI, template generation, motion graphics, design automation, and AI-assisted editing.
View Dataset ALPHA CHANNEL VIDEO DATASETVideo Alpha Channel Dataset
Access 26,880 curated 4K and HD alpha channel video clips with human-authored metadata for video generation, visual effects, compositing, multimodal AI, and model training.
View Dataset