On this page
Separate audio identity from consent evidence
The source dates shown in this guide identify documentation checks on 22 October 2026, not original publication dates. Follow each citation for the timing and provenance vocabulary being discussed. Begin with a controlled audio identifier, digest or owner-approved reference, duration, channel description, derivative relationship, and handling boundary. Do not paste raw speech, names, phone numbers, private paths, or customer recordings into a public worksheet.
Consent-related evidence is a record of what was supplied and how it is linked, not a conclusion that a legal requirement was met. Keep source document identifier, collection context, stated scope, version, time, issuing or collecting role, and evidence location separate from an owner or counsel decision. Mark unavailable fields as unknown rather than inferring permission from the existence of a file.
| Evidence layer | Synthetic field | Do not infer |
|---|---|---|
| Audio entity | synthetic-audio-07 and digest | Who may use the speech |
| Transcript entity | synthetic-transcript-07-v2 | That words match the audio |
| Collection activity | recorded-event-07 | That the event was authorized |
| Consent record | consent-record-07, scope unknown | That a legal standard is satisfied |
| Agent or role | synthetic collection operator | Identity of a real speaker |
| Decision | owner review pending | Approval, clearance, or license |
Check transcript cues against the media timeline
The World Wide Web Consortium, (opens in a new tab) specification describes cues with start and end timestamps and allows cues to overlap. Record the transcript format, timestamp precision, media duration used for review, and whether cue boundaries were produced by a human, an automated process, or an unknown stage. Overlap is an observation; it is not automatically an error because multiple speakers or displayed tracks may overlap.
Synthetic cue ledger
cue_id | start | end | speaker_id | audio_ref | evidence
cue-01 | 00:00:01.200 | 00:00:03.100 | speaker-a | synthetic-audio-07 | aligned-window
cue-02 | 00:00:02.900 | 00:00:04.000 | speaker-b | synthetic-audio-07 | overlap-review
cue-03 | 00:00:04.000 | 00:00:04.000 | unknown | synthetic-audio-07 | zero-duration-hold
cue-04 | 00:00:05.000 | 00:00:06.250 | speaker-a | synthetic-audio-08 | wrong-audio-link| Cue check | Evidence | Disposition |
|---|---|---|
| Start before end | Normalized cue interval | Hold zero or reversed interval |
| Within reviewed duration | Media duration and cue endpoint | Record out-of-range cue |
| Audio reference matches | Audio id or approved digest | Hold wrong-derivative link |
| Speaker mapping stated | Stable synthetic or owner-approved id | Keep speaker unknown |
| Overlap explained | Concurrent speaker or track note | Do not silently merge cues |
| Text source retained | Transcript version and activity | Do not overwrite prior transcript |
A cue ending exactly where the next cue starts may be contiguous. A zero-duration or reversed cue needs a disposition, not a guessed correction. Keep cue text out of broad reports unless the owner has approved the handling boundary; a stable cue id and timing finding often suffices.
Model derivation and collection as an evidence chain
Use PROV-O as a vocabulary for the chain: the audio and transcript are entities; recording and transcription are activities; a collection role, annotator, or system is an agent. Record that a transcript was derived from a specific audio entity and generated by a named activity. This makes missing links visible without claiming that provenance alone proves accuracy or authority.
| PROV-style relation | Synthetic record | Review question |
|---|---|---|
| used | transcription activity used synthetic-audio-07 | Is the input identity stable? |
| wasGeneratedBy | transcript-v2 generated by activity-22 | Is the producing step known? |
| wasDerivedFrom | transcript-v2 from audio-07 | Is the derivative relationship supported? |
| wasAssociatedWith | activity-22 associated with annotator role | Is the role evidence available? |
| wasAttributedTo | transcript attributed to owner role | Is attribution distinct from approval? |
Store the event time, tool or workflow version, and evidence locator separately. If a source record is withheld, report that the link is unavailable. Do not put a raw transcript, consent form, speaker name, or customer identifier into a public synthetic fixture.
Use a consent-evidence worksheet with unknown states
The worksheet below intentionally uses evidence states rather than legal labels. An owner can fill it with a controlled reference and have the appropriate privacy, compliance, or counsel process decide what the evidence means. The guide does not decide that question.
| Field | Synthetic value | Allowed disposition |
|---|---|---|
| Record reference | consent-record-07 | supplied, withheld, or unavailable |
| Collection context | stated context only | documented, partial, or unknown |
| Scope text | not reproduced here | present, partial, or unknown |
| Version and time | 2026-10-20 / v1 | linked, conflicting, or missing |
| Audio linkage | audio-07 digest reference | matches, mismatch, or unverified |
| Speaker linkage | speaker-a synthetic id | mapped, unknown, or not provided |
| Decision owner | owner role pending | pending, recorded, or escalated |
Keep transcript findings and legal decisions separate
This review can show whether a transcript is linked to an identified audio artifact, whether cues have bounded timing evidence, and whether collection and derivative records are traceable. It cannot establish speaker identity, truth of transcript text, privacy status, permission, legal compliance, ownership, or suitability for a downstream use.
Pair this guide with the data provenance log for artifact history and transcript alignment review for cue and speaker observations. After supported non-health scope, authority to describe the collection, and a controlled evidence packet are established, the qualified appraisal request path may be a next preparation step. It does not inspect recordings, decide consent, create a quote, match a buyer, issue a legal opinion, or grant a license.
Takeaways
- Keep audio identity, transcript identity, collection activity, consent evidence, and owner decisions as separate records.
- Use cue ids and timing observations instead of copying raw speech into a public report.
- Record overlaps and unknown speaker links without forcing a merge or identity.
- Use provenance relations to expose missing derivation links, not to claim accuracy or authorization.
- Treat consent-related fields as evidence inputs for the appropriate process, never as a legal conclusion.