Format & structure
Confirm file types, duration, channel layout and source references against the proposed sample.
Technical delivery
Understand the data
before you use it.
A useful collection connects its source files, representations and permissions. Agree the specification, inspect a representative sample, and define what an accepted delivery contains.
Connected assets
One source. A clear delivery record.
01Source session
A stable identifier links the original recording to its components.
02Media assets
Included audio, video and related files are identified individually.
03Representations
Transcripts and annotations retain their source references.
04Delivery record
The manifest identifies the agreed files and supporting records.
This diagram explains a delivery structure. Components and annotations depend on the selected collection and agreement.
Source media and annotations are different deliverables.
An original recording does not automatically include an accurate transcript, speaker segmentation, action labels or calibrated sensor streams. We identify existing representations, additional preparation and acceptance criteria in the project scope.
Find requirements for your model task ↗Review together
Make quality specific to the task.
Coverage & context
Review the languages, environments, perspectives and sequence boundaries your task needs.
Annotation scope
Define conventions, timestamp requirements and review criteria for the annotations actually included.
Known exceptions
Identify exclusions, gaps and any preparation still needed before agreeing the final delivery.
Transfer & acceptance
Agree manifest fields, integrity checks and the process for resolving missing or unusable assets.
Inspectable documentation
Know what each field means.
The public manifest demonstrates the schema using fictional identifiers. Project-specific files and sensitive records are shared within the agreed review process.
| Field | Meaning | Review purpose |
|---|---|---|
content_id | Connects assets from the same source session. | Stable source-to-asset relationship |
asset_id / role | Identifies a particular audio, video or transcript asset. | Which components are included |
format / duration_seconds | Describes the asset’s representation and length. | Compatibility with the agreed specification |
sha256 | File checksum in an agreed delivery manifest. | Integrity during transfer |
rights_ref | Reference to the applicable rights documentation. | Connection to the permitted use |
consent_status / privacy_status | Records review status for the proposed use. | Outstanding conditions and exclusions |
A shared acceptance standard
Agree what “ready” means.
Use a representative sample to define acceptable formats, context and annotations. Record required checks and unresolved conditions before the collection is prepared or transferred.
Request a collection sample ↗Review a proposed delivery
- Selected source scope and exclusions
- Files, formats and annotations included
- Stable identifiers and integrity requirements
- Permissions, privacy and consent conditions
- Acceptance process and delivery milestones