# Audio Transcription
An Audio Transcription preserves speech from a specific audio or video source with timestamps, speaker labels, uncertainty markers, redactions, and source provenance. The audio file remains the primary source; the transcript is a derived navigational and analytical layer.
## Required distinctions
- Recording date versus transcription date
- Certain, probable, and unknown speakers
- Exact transcript versus editorial description
- Audible words versus inferred context
- Complete recording versus excerpt, edit, or compilation
- File metadata versus independently established event time
## Minimum record
Record the source filename as plain provenance unless the media exists in the vault; duration; recording and transcription dates; file hash when available; participants and speaker confidence; location confidence; transcript completeness; redactions; time-coded excerpts; related events, people, and topics; and chain-of-custody notes.
Personal phone numbers and password-equivalent information remain redacted. No credential, account, telephone number, or service discovered in a recording may be tested or contacted.
## Naming and navigation
Use `Audio - YYYY-MM-DD - Short Description.md` when the recording date is known, or `Audio - Undated - Short Description.md` otherwise. See [[Index - Audio Transcriptions]].