Transcribing a Parent's Recorded Stories: Which Fields to Preserve

The recording remains the primary source

A transcription converts spoken language into written text. It can make an old voice recording searchable, easier to quote and more accessible to someone who cannot conveniently listen to the entire file. It also changes the form of the material. A transcript cannot preserve the original voice, pauses, background sounds, overlapping speech, emphasis or every detail of pronunciation. Those qualities remain in the audio recording.

For that reason, retain the original audio file as a separate preservation object. Do not overwrite it with a cleaned, shortened or noise-reduced version. If an improved listening copy is useful, save it as a derivative with a distinct filename. The transcript should identify the recording from which it was produced, so that a future reader can move between the written text and the original sound.

• Original audio recording.
• Unmodified preservation copy.
• Optional listening or access copy.
• Transcript linked to the original filename.
• Notes describing any edits or processing.

Prepare old voice recordings before transcription

Before using an oral history transcription tool, establish what kind of recording is available. Voice memos may be stored as M4A, AAC or another compressed format. Older recorders may produce WAV, MP3 or proprietary files. A recording copied from a cassette or other analogue medium may also contain transfer noise, uneven volume or gaps. Record the source and any known transfer history because these details can affect how confidently individual words can be understood.

Give the audio a stable filename before creating a transcript. For example, 1998_UnknownDate_Interview_001.m4a indicates a recording from 1998 with an uncertain exact date. Add a separate catalogue note identifying the speaker, recording circumstances and any known participants. Avoid changing the original recording's date metadata merely to make the filename appear more complete.

• Record the existing filename and file format.
• Identify the device or medium where known.
• Record the recording date or estimated period.
• Preserve an unchanged copy before processing.
• Note any transfer, conversion or audio enhancement.

The transcript fields worth retaining

A transcript needs more than the words alone. At minimum, identify the source recording, the transcript's filename, the language and the person or people speaking. Add timestamps at regular intervals or at meaningful changes in topic. If a recording contains several speakers, use consistent labels throughout. These fields allow someone to interpret the document without needing to remember how it was created.

For example, a transcript might begin with the recording reference, approximate recording date, speaker names and transcription date. Its first timestamp might read [00:00:42] SPEAKER 1: 'The house was beside the old railway line.' A timestamp identifies a position in the recording, not the date on which the statement was made. Keep those two kinds of date separate.

• Recording filename or stable identifier.
• Recording date and date precision.
• Speaker labels and known identities.
• Language and any relevant dialect information.
• Timestamps linked to the original audio.
• Transcription date and method.

Timestamps and speaker labels preserve context

Timestamps are among the most useful fields in a voice memo to text archive. Without them, a reader may have to search through an entire recording to locate one sentence. A timestamp every 30 to 60 seconds is a practical starting point for a straightforward interview, although shorter intervals may be appropriate when speakers change frequently or the subject contains many distinct events. The interval should suit the recording rather than follow a rigid rule.

Speaker labels need similar care. If the speaker is known, use a consistent name or role, supported by the recording context or accompanying notes. If the identity is uncertain, use SPEAKER 1 or an equivalent neutral label. A transcription system may assign labels automatically, but those labels can change when speakers interrupt one another or speak over background noise. Preserve the label as a working identifier until the identity has been established.

• Use elapsed recording time in HH:MM:SS format for longer files.
• Keep timestamps attached to the passage they identify.
• Use the same speaker label throughout one transcript.
• Mark overlapping or indistinct speech where relevant.
• Record uncertain speaker identities separately.

Mark uncertainty instead of inventing missing words

Automated transcription can produce plausible sentences from unclear speech. This is useful for preparing a draft, but a plausible sentence is not necessarily an accurate record. Names, place names, dates, local expressions and unfamiliar technical terms deserve particular attention. A recording of a family member discussing a village might produce a familiar-sounding alternative name when the original pronunciation is difficult to recognise.

Use a consistent notation for uncertainty. For example, [unclear 00:04:18] identifies a passage that cannot be confidently understood, while [possibly 'Ravenswood'] records a tentative reading. Keep editorial additions in square brackets so that they cannot be mistaken for the speaker's words. If a transcript is corrected against the audio, record that editorial intervention in its notes rather than presenting the text as an untouched automatic transcript.

• Use [unclear] for speech that cannot be understood.
• Use [inaudible] when speech is present but cannot be recovered.
• Mark uncertain names and places explicitly.
• Distinguish editorial explanations from spoken words.
• Do not silently complete sentences that are interrupted or unfinished.

Choose transcript formats that can travel

A plain TXT file is a useful basic transcript because it can be opened by a wide range of text editors. DOCX can preserve more structured formatting, headings and tables, but depends on compatible office software for full presentation. PDF can provide a stable reading copy, although it is less convenient for editing or extracting structured information. A practical archive can retain a TXT master alongside a formatted reading copy where the distinction is useful.

For transcripts with precise timing, WebVTT or SRT can preserve timed text in a structured form. These formats are especially useful when subtitles or synchronised playback are required. They are not substitutes for a complete descriptive transcript in every situation. If a transcription service supplies a proprietary project file, retain it only as an additional copy and export the text and timing information into a more widely readable format.

• TXT: simple, portable transcript text.
• DOCX: formatted text with editable structure.
• PDF: convenient fixed-layout reading copy.
• WebVTT or SRT: timed text for synchronised playback.
• Original audio: retained independently of all transcript formats.

A practical handover structure for recorded stories

Keep each recording and its transcript together in a clearly named folder. For example, a folder called 1998_Family_Stories might contain the original M4A file, a TXT transcript, a formatted PDF reading copy and a short README. The README should explain the filename convention, identify the speakers where known and describe any uncertainty in the recording dates. If there are several recordings, maintain a simple index connecting each title to its filename and duration.

AI transcription features may also supply language tags, automatic summaries and topic labels. Preserve these as separate descriptive fields and identify them as generated or editorial material. A summary can help someone decide which recording to open, but it should never replace the complete transcript or audio. Start with one recording, preserve its original file, create a transcript with timestamps and speaker labels, and add its reference to the collection index.

• Store the original recording and transcript in the same collection.
• Keep a short README describing the folder.
• Include an index of filenames, speakers and durations.
• Identify automatically generated descriptions.
• Record the date and method of any later transcription correction.

Questions readers ask