What it costs to transcribe a podcast archive
Human transcription runs roughly £1–£2 per audio minute, so £60–£120 an hour — which for a sixty-hour archive is £3,600–£7,200. Automated is a small fraction of that. The more useful answer is that you should not transcribe most of an archive at all: work out what the book is first, then transcribe what serves it.
What it costs, by route
Prices dated 9 September 2026. The bottom row is the one worth reading twice.
| Route | Per audio hour | 60-hour archive | When it's right |
|---|---|---|---|
| Human transcription | £60–£120 | £3,600–£7,200 | Poor audio, heavy accents, overlapping speakers, specialist terms |
| Automated, then human check | £20–£45 | £1,200–£2,700 | Decent audio where accuracy matters and budget is limited |
| Automated only | A few pounds | Low hundreds | Clear single-speaker audio; expect proper-noun errors |
| Selective — analyse first, transcribe 15% | As above, on 9 hours | £540–£1,080 human | Almost always. This is the recommendation. |
A separate free preview: Find the book already hiding in your podcast
A separate podcast-to-book preview based on public feed details; it does not transcribe audio, listen to episodes or process private recordings. Paste your public RSS feed. You get the strongest book angle in your archive, who it is for, the three to five themes you keep returning to, a possible chapter spine, the five to ten episodes worth starting with, and what the book is still missing.
A separate podcast-to-book preview based on public feed details; it does not transcribe audio, listen to episodes or process private recordings. Built from your public feed only. No archive upload, no transcription, and nothing is written or cleaned up for you.
Why the last row is the right answer
You will use perhaps 10–15% of an archive in a book. Transcribing the other 85% produces a document nobody reads, at full price.
So the sequence that saves the money: analyse the whole archive thematically, choose the through-line, then transcribe only what serves it.
On a sixty-hour archive that is the difference between £3,600–£7,200 and roughly £540–£1,080 for human work on the material you actually need — with the hardest recordings sent to a person and the rest done automatically.
Where automated transcription actually fails
It is now excellent on clear single-speaker audio in a familiar accent, and noticeably worse on four things.
Proper nouns. The biggest practical problem for a book: names, places, companies, technical terms come back wrong, consistently, and propagate through 200 pages if not caught.
Overlapping speakers. Interview shows with genuine conversation suffer most.
Strong regional accents, in both directions.
Poor rooms. Echo and background noise degrade accuracy sharply.
The cheapest single improvement is a name list prepared before you start — every person, place and recurring term. It improves a transcript more than any other intervention at any price.
What you are paying a human for
Four to six hours of careful work per hour of audio on decent recordings, and considerably more on difficult ones. That is the whole explanation of the price.
What you get: accurate proper nouns, reliable speaker attribution, sensible handling of crosstalk, and judgement about what a mumbled phrase actually was.
For the recordings your book genuinely depends on — the interview with the person who has since died, the episode everything hinges on — that is worth paying for.
About Your Book Pro
Your Book Pro is a book-writing platform built in Britain. Start with an idea, notes, half a draft, a podcast, or a finished manuscript, and use our purpose-built AI tools for planning, feedback, editing, proofreading, cover designs, and to create print-ready files. Writing is totally free: you can type, upload or speak your book as much as you like, and use only the editing tools you need with Plus, or go unlimited with Pro.
Produce world-class drafts at the fraction of the cost of hiring a human editorial team.
Every account starts with 1,000 welcome credits, and a plain export of your own words is free.
Two things change the arithmetic. Themes from Bulk Audio reads the whole archive before transcription and tells you what the material is about, so you transcribe a fraction rather than all of it — which is where the real saving is.
And Transcription produces clean verbatim with speaker labels and timestamps, checked against your Book Dictionary, so the names and recurring terms you have listed come back spelled correctly rather than differently on every page. It runs on credits, and every account starts with 1,000 that never expire.
What still favours a human: heavy accents, several people talking over each other, poor rooms, and legal or medical terminology. Budget £60–£120 an audio hour for those specific files and use the tools for the rest. That hybrid is what most people should do, and we would rather say so than sell you the whole archive.
Upload your archive
See what's in it first
Email me my free podcast summaryQuestions people ask
- How much does it cost to transcribe a podcast?
- Human transcription is roughly £1–£2 per audio minute, so £60–£120 an hour. Automated is a small fraction, with accuracy depending heavily on your audio.
- Should I transcribe my whole archive?
- Almost never. Analyse thematically first and transcribe the 10–15% that serves the book. It is the single biggest saving available here.
- Is automated transcription accurate enough for a book?
- For clear audio, close — with the exception of proper nouns, which need a prepared name list and a check. For difficult audio, no.
- What's the cheapest reliable approach?
- Analyse first, transcribe selectively, use automated transcription for clear recordings and a human for the difficult ones that matter.
- Do I need timestamps?
- Yes. When you are assembling a book from sixty hours, the timestamp is how you go back and check what was actually said — and you will need to more often than you expect.