CUSTOMMULTIMODALPAIRED DATA
Multimodal Paired Capture Sets
Paired multimodal datasets linking video, audio, screen, text, or image-derived records through shared file IDs, cross-modal manifests, metadata schemas, and QA fields.
CUSTOMANY MODALITYBUYER-SCOPED
Buyer-Scoped Custom Collection
Custom data collection format for buyers who need a dataset designed around a specific model objective, modality mix, metadata schema, QA workflow, and delivery structure.
Buyer-Scoped Custom Collection is built for teams whose dataset requirements do not fit a standard collection category. The format is designed around the buyer’s model objective, target modality, collection rules, contributor requirements, annotation needs, metadata fields, QA criteria, and delivery format.
Each collection is documented through a dataset specification process before collection begins.
Structured collection
Buyer-Scoped Custom Collections are built from a dataset specification rather than a generic task template.
Metadata depth
The format can support transcript alignment, video-step alignment, action-event alignment, screen-event alignment, cross-modal…
Delivery-ready package
Includes system, content, metadata reviews, plus dataset manifest, CSV catalog, consent status metadata, QA summary, delivery…
Request sample access
Share your target modality, model objective, metadata schema, annotation needs, QA requirements, and delivery format. We’ll reply with a technical scope for the collection.
Key highlights
- Dataset specification built around buyer-defined model objectives and pipeline requirements.
- Modality-specific capture rules for audio, video, screen recordings, text-linked data, or approved multimodal combinations.
- Custom metadata schema with task IDs, file IDs, annotation fields, QA status fields, and delivery batch structure.
Dataset contents
Dataset files
Buyer-scoped files · Task definitions · Capture instructions · Accepted file type rules
Task and capture definitions
Contributor submission metadata · Modality-specific metadata · Annotation schema where scoped · QA status metadata
Technical specifications
Use cases
- Custom model evaluation
- Buyer-specific benchmark creation
- Multimodal assistant testing
- Specialized speech collection
- Screen workflow collection
- Embodied AI data collection
- Dataset prototyping
- Pipeline-specific training data collection
Dataset workflow
1Collection
Buyer-Scoped Custom Collections are built from a dataset specification rather than a generic task template. Byser defines the collection workflow, contributor instructions, review fields, and delivery schema around the buyer’s actual pipeline requirements.
2Review
System check · Content review · Metadata review · Annotation consistency review · Consent status review · PII review where applicable · Delivery package review · Final approval
3Delivery
Structured files by modality · Dataset specification summary · Metadata CSV · Annotation files where scoped · Dataset manifest JSON · CSV file catalog · QA summary · Consent status metadata · Delivery notes
Quality assurance
Related dataset formats
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.
CUSTOMAUDIODIALOGUE
Natural English Dialogue Sets
Multi-speaker English dialogue sets with speaker-labeled transcripts, timestamped turns, accent/session metadata, optional separated-channel audio, and buyer-scoped annotations for ASR, diarization, and voice-agent evaluation.
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.
CUSTOMIMAGEPRODUCT RECOGNITION
Product and Packaging Image Sets
Product and packaging image sets with multi-angle capture, lighting metadata, package-side labels, category fields, resolution checks, QA status, and optional bounding-box or OCR annotations.
CUSTOMIMAGEOCR
Real-World Text and Signage Photos
Scene-text image sets with language/script metadata, text-region boxes, ground-truth transcription, scene category labels, QA fields, and manifest records for OCR and visual translation.