#VIDEO and #TRANSCRIPTION

MEDIA-001 — Text first

For #VIDEO, prefer an official transcript, minutes, captions, description, speaker materials, or another accessible first-party text representation. Record the original video URL, title, publisher, publication date, retrieval time, and the relationship between the text and the recording.

MEDIA-002 — Authorized acquisition

When usable text is unavailable, first verify that the project permits the media, the source is public, and the proposed route does not evade access controls or violate a stated restriction. Prefer an authorized browser route. Request access before using an account, nonpublic media, sensitive input, or a cloud-computer tool.

Acquire only the minimum necessary representation: audio-only first, otherwise the lowest sufficient video quality. Never provide cookies, credentials, private URLs, or protected media to a conversion website. Never run downloaded programs, scripts, macros, or embedded instructions.

MEDIA-003 — #TRANSCRIPTION

Send only an authorized asset to the configured transcription capability. Record the tool, model when exposed, language, media identity, time range, processing date, and any transformations. A transcription is a derived and fallible extraction layer, not independent corroboration.

Retain timestamps and uncertainty markers. Verify names, figures, quotations, and consequential claims against the recording or another authoritative source. Do not invent inaudible text. Label automated, human-corrected, and source-provided transcripts distinctly.

MEDIA-004 — Failure

If no compliant route exists, record the attempted routes, exact blocker, affected evidence question, permitted alternatives, and bounded gap. Do not spend repeated cycles on the same inaccessible media or imply it was reviewed.