CUSTOMSCREENCOMPUTER-USE
Software Task Screenflows
Screen workflow recordings with task objectives, OS/app metadata, cursor activity, step boundaries, completion status, and action logs for computer-use model evaluation.
CUSTOMMULTIMODALPAIRED DATA
Multimodal Paired Capture Sets
Paired data collection format for vision-language models, audio-visual learning, screen-and-voice workflows, multimodal instruction following, and cross-modal evaluation.
Multimodal Paired Capture Sets are built for teams that need files from multiple modalities linked through a shared task, timestamp, contributor session, or model objective. The format is designed for use cases where isolated audio, video, screen recordings, or text are not enough.
Each collection is structured around a modality pair or approved modality combination, such as screen plus voice narration, video plus audio, first-person video plus transcript, or image-derived capture plus text description.
Structured collection
Multimodal Paired Capture Sets are scoped around a model objective and a defined modality relationship.
Metadata depth
The format supports file-level pairing, segment-level pairing, timestamp-level alignment, transcript-video alignment,…
Delivery-ready package
Includes modality completeness, pairing consistency, alignment reviews, plus dataset manifest, CSV catalog, consent status…
Request sample access
Share your target modalities, pairing logic, alignment requirements, annotation needs, and evaluation goals. We’ll reply with a technical scope for the collection.
Key highlights
- Paired file structure connecting two or more modalities through shared session IDs or task IDs.
- Cross-modal manifest records linking audio, video, screen, transcript, annotation, and metadata paths.
- Support for screen plus narration, video plus audio, POV video plus transcript, and buyer-scoped modality combinations.
Dataset contents
Component files
Paired audio files where scoped · Paired video files where scoped · Screen recordings where scoped · Transcript files where scoped
Pairing and transcript records
Timestamp alignment fields where scoped · Text descriptions where scoped · Shared session IDs · Shared task IDs
Cross-modal metadata
Cross-modal pairing metadata · Modality-specific metadata
Technical specifications
Use cases
- Vision-language model evaluation
- Audio-visual learning
- Multimodal instruction following
- Screen-and-voice workflow analysis
- Video-transcript alignment
- Cross-modal retrieval evaluation
- Embodied AI data pairing
- Multimodal benchmark creation
Dataset workflow
1Collection
Multimodal Paired Capture Sets are scoped around a model objective and a defined modality relationship. Each task specifies what modalities are captured, how files are paired, what alignment is required, and how the delivery package should be organized.
2Review
Modality completeness review · Pairing consistency review · Alignment review · Metadata review · Annotation consistency review · Consent status review · Delivery approval
3Delivery
Component files by modality · Pairing metadata CSV · Cross-modal manifest JSON · Transcript files where scoped · Annotation files where scoped · CSV file catalog · QA summary · Consent status metadata · Delivery notes
Quality assurance
Related dataset formats
CUSTOMSCREENWEB WORKFLOWS
Browser Navigation and Web Task Flows
Browser task recordings with URL context, click and scroll events, form interaction metadata, page categories, task outcomes, and action logs for web-agent evaluation.
CUSTOMSCREENNARRATED WORKFLOWS
Guided Software Walkthrough Sessions
Narrated software walkthroughs with screen recordings, aligned narration transcripts, step markers, app context, cursor flow, and task metadata for procedural software understanding.
CUSTOMVIDEOFIRST-PERSON
First-Person Activity Clips
Wearable-style first-person activity clips with task labels, step metadata, camera perspective, hands-visible review, environment fields, and optional narration for embodied AI evaluation.
CUSTOMAUDIOACOUSTIC SCENES
Acoustic Environment Sets
Ambient audio sets with scene labels, acoustic context metadata, speech screening, device fields, noise-level estimates, and QA status for acoustic scene and robustness evaluation.
CUSTOMANY MODALITYBUYER-SCOPED
Buyer-Scoped Custom Collection
Fully scoped custom data collection with buyer-defined modality, schema, metadata fields, annotation layers, QA criteria, manifest structure, delivery package, and collection rules.