Work in passes, and do them in order: cut the filler, impose paragraphs, fix the circling, add what the eye needs that the ear didn't, then read it aloud. Do not try to make it sound written on the first pass — that is how the spoken voice, which is the best thing in the draft, gets erased.
How to edit a spoken draft into prose
The five passes, in order
One pass at a time, on the whole chapter, before starting the next. Doing them together is why transcripts end up sounding like nobody.
- PASS 1 — Filler. Cut um, er, you know, sort of, I mean, basically. Cut false starts that carry no meaning.
- PASS 1 — Cut the throat-clearing at the start of answers: 'Well, that's a good question, I suppose...'
- PASS 2 — Paragraphs. Speech has none. Break at each shift of subject, and again wherever the eye needs rest.
- PASS 2 — Split the sentences that ran on for ninety words because you didn't need to breathe.
- PASS 3 — Circling. Speech says a thing three ways. Keep the best one and delete the other two.
- PASS 3 — Move material that arrived out of order into the place it belongs.
- PASS 4 — Add what the eye needs: who is speaking, where this is, how much time has passed.
- PASS 4 — Add the detail you gestured at aloud. 'That big place on the corner' needs a name on the page.
- PASS 5 — Read the whole thing aloud. Anything you stumble on, fix. Anything that no longer sounds like you, restore.
- Throughout: keep the speaker's word choices, dialect and grammar unless they obscure meaning.
- Throughout: keep one or two of the verbal habits. A wholly smoothed transcript sounds like nobody.
A cleaned-up chapter
Send a transcript or a chapter and we'll email it back with the filler gone and paragraphs in — wording untouched — plus a short note on the habits worth keeping and the ones worth cutting.
Why speech and prose are different jobs
A listener has your tone, your pauses, your face, and no ability to go back. A reader has none of the first three and total control of the fourth.
So speech compensates in ways prose does not need: it repeats to make sure you caught it, it uses filler to buy thinking time, it circles back, and it gestures at things — that big place on the corner — that a listener can accept vaguely and a reader cannot.
Editing a transcript is largely removing the compensations that are no longer needed and adding the ones the page requires.
The single biggest mistake
Trying to make it sound written on the first pass.
The instinct is to tidy as you go — smooth the grammar, lift the vocabulary, straighten the syntax. The result is prose that reads competently and sounds like nobody, and the thing you have lost was the reason for recording it.
Your grandmother's odd word order is not an error. The phrase she always uses is not filler. The way she starts a story from the wrong end is often the best structural instinct in the chapter.
Cut the noise. Leave the person.
Paragraphing is where most of the improvement is
A transcript is a wall. Break it and it becomes readable, before you have changed a single word.
Break at each shift of subject, at each shift in time, and whenever the block on the page is getting heavy. Short paragraphs are not a failing in a spoken book; they mirror the rhythm of how it was said.
Do this pass alone and reread. Most people find the transcript is far better than they thought — it was drowning, not weak.
Keeping a couple of the tics on purpose
Every speaker has habits: a phrase they return to, a way of qualifying, a particular joke shape. Remove all of them and the voice goes flat.
The working rule: keep the two or three that are characteristic, cut the ones that are merely frequent. If your father says anyway at the start of every second sentence, keep it three times in a chapter and cut the rest. It reads as him. Twenty-eight times reads as a transcript.
This is a judgement call and it is the part that cannot be automated, which is exactly why it is worth your attention rather than the filler-cutting.
What Your Book Pro does here
Your Book Pro is a British book-writing platform. You bring a book — an idea, a folder of notes, half a draft, or a finished manuscript written somewhere else — and it gives you the editorial process a publisher would run: a plan, structured feedback, a simulated beta read, a line edit, a proofread, a cover, a blurb and a print-ready file. You can type it, upload it or speak it. The tools sit in one place and work on the same manuscript, so the feedback compounds instead of arriving as twelve disconnected opinions.
This is the workflow the platform was built around, and the passes are kept separate on purpose. Transcription produces clean verbatim first — you see the real words before anything is touched. Light Edit then does passes one and two: filler and repetition out, paragraphing in, without rewriting your wording. What comes back is your sentences, breathing.
Passes three, four and five are judgement, and they are yours. Ask the Editor is there while you do them, for the specific question — does this still sound like her, or have I flattened it?
If you have a lot of material, Themes from Bulk Audio reads across all of it and tells you what the book is, which is the question you cannot answer chapter by chapter.
And if the writing itself is the point — literary memoir, essays — a human line editor will do pass three better than any tool, because knowing which of three versions of a sentence to keep is craft. Do passes one and two here first; it makes their work faster and their quote smaller.
Clean up a chapter
Upload recordings or a transcript
Email me the chapterQuestions people ask
- How long does it take to edit an hour of transcript?
- Roughly three to six hours by hand for a first usable draft, depending on the recording. The mechanical passes are the bulk of that, which is why removing them first changes the arithmetic.
- Should I edit the transcript or rewrite from it?
- Edit, for memoir and family books — the voice is the value. Rewrite, for business books where clarity matters more than character. Decide before you start rather than drifting between the two.
- What about grammar in spoken drafts?
- Correct it where it obscures meaning, keep it where it is characteristic. Regional grammar in a family memoir is not an error to be fixed; it is the reason the book exists.
- How do I handle two people talking?
- Decide whether the book is an interview or a narrative. As an interview, keep the questions and label the speakers. As narrative, remove your questions and stitch the answers — which needs pass four, because the questions were carrying the structure.
- Can I dictate the edits too?
- Yes, and many people find it faster. Talk through what needs changing, then apply it. There is no requirement to type at any stage.