Transcription Foundations: Answers to the Question We Hear Most

The question we hear most often in the transcription module is simple: how clean does the text have to be before I start timing it? Beginners worry that they are moving too early or too late, and early autumn brings a wave of new learners with recorded talks from summer events waiting in a folder.

Our answer is that timing should begin only when the words are settled. Timing a rough draft means redoing the work every time a sentence changes, because line breaks and start and end points depend on the final wording. A clean transcript first saves hours later.

That does not mean perfection. It means every sentence is complete, every speaker's words are in the right place, and the doubtful passages are marked rather than guessed. This note explains what settled looks like in practice, and how to get there without strain.

The rest of this note follows the order in which learners usually raise their worries: what clean means, how to treat terminology, what to do with muddy audio, and when to stop.

A clean transcript reads the way the speaker meant to be understood. It removes false starts, repeated filler words and stumbles that carry no meaning, while keeping every word that changes what was said. Marcus calls this the test of the quoted sentence: if someone repeated your line aloud to the speaker, would they recognise it as their own?

Punctuation matters more than most beginners expect. A comma in the wrong place can change who did what, and a missing question mark turns a doubt into a statement. Read each paragraph aloud once before moving on.

Keep one consistent set of choices throughout the file. Decide early whether numbers appear as digits or words, how you write out acronyms on first use, and whether you keep spoken contractions. Write these choices down in a short list at the top of your working document.

Consistency is what reviewers notice first, long before they notice a single missed word.

• Remove stumbles and repeats that add no meaning
• Keep every word that changes the sense
• Read each paragraph aloud once
• Write down your style choices before you begin

Talks and interviews are full of names, product terms and specialist vocabulary that no ear can guess. The temptation is to type what it sounds like and move on. Resist it. A wrong term in a caption is worse than a gap, because viewers trust it.

Ask for a glossary before you start. Even a half-page list from the person who supplied the recording saves most of the guessing: names of speakers, place names, programme titles and any technical words. If no list exists, build your own as you go and mark each uncertain entry with a flag so it can be checked later.

When you cannot confirm a term, search for the most likely spelling in a trustworthy source, then mark it in your notes as unverified. Never invent a spelling to make the page look finished.

In class we practise this with sample interviews where three terms are deliberately unfamiliar, so learners get used to pausing rather than guessing.

• Ask for a glossary of names and terms first
• Flag every uncertain spelling as you type
• Never invent a spelling to fill a gap

Poor audio is the second most common worry. Room echo, a speaker turning away from the microphone, a clinking glass at the wrong moment: all of these appear in event recordings, and none of them is a sign that you are doing something wrong.

Start by listening to the difficult passage three times at normal speed, then once slowly. Headphones help far more than laptop speakers. Often the sentence around the unclear word gives you the meaning, and the word itself follows from context.

If a passage stays unclear after those attempts, mark it with a short tag such as an unclear note and its approximate position. Do not spend twenty minutes on four seconds of sound. Return to it at the end of the file, when you know the speaker's habits better and the topic is familiar.

Elena's advice from the speaker and sound notes module applies here too: your ear improves with each minute of one voice, so difficult passages are easier on a second pass.

• Listen three times at normal speed, then once slowly
• Use headphones rather than laptop speakers
• Tag what stays unclear and come back at the end

Work in passes rather than in one long sitting. The first pass is a rough typing of everything you can hear, without stopping to fix anything. The second pass cleans wording, punctuation and style. The third pass handles the flagged terms and unclear passages.

Short sessions give better results than marathons. Many learners find that forty minutes of focused work followed by a break produces fewer errors than two hours straight, and the errors that do appear are easier to spot with fresh eyes.

Keep the original recording and your working file together, named clearly with the date and the version. When a reviewer asks why a sentence reads a certain way, you can return to the exact moment in the audio and answer from evidence instead of memory.

These habits are not complicated, but they are what separate a transcript that moves smoothly into timing from one that sends you back to the beginning.

• First pass: type everything, fix nothing
• Second pass: clean wording and punctuation
• Third pass: resolve flags and unclear passages

Learners ask how they will know. We give them a short checklist. Every sentence is complete and punctuated. Every speaker change is marked. Every flagged term is either confirmed or clearly noted as unverified. Your style choices have been applied from the first line to the last.

One more check helps: listen to a few random minutes with the text beside you and no pause button. If you can follow the text without losing your place, the transcript is close to ready. If you keep stopping to correct, go back through one more pass.

Once those checks are done, the transcript moves into timing and subtitle sync, where the text is broken into readable caption lines. That stage is covered in its own part of the curriculum, and it goes much faster on a transcript you trust.

Settled text is the real answer to the question we started with.