01 / CRUDE
Crude
Your recording arrives as it arrived. A bad room, a phone on the table, three people talking over each other. Nothing is discarded at this stage, and nothing is trusted either.
Transcribe Refinery, professional transcription and document refinement
Raw audio is a crude material.
Where it gets hard, a machine guesses.
We listen again. Every time.
Recorded speech refined into a written record that is faithful to the audio, clear for the reader, and handled with discretion.
Send us 5 minutes, free
Standard, Express or Urgent
Transcribe Refinery
Every word, decided.
Recorded speech refined into a clear, accurate written record. Faithful to the audio, and handled with discretion.
Send us 5 minutes, free
98%
The accuracy the machines advertise, measured on clean test audio
The gap
Even at 98 percent, a thousand words still carries twenty errors.
That is their own best case, and it is measured on clean recordings made with one speaker, no crosstalk and no room noise. Independent testing puts real audio a long way below it. One study of the same system measured 92 percent on a headset, 78 percent in a meeting room, and 65 percent on a mobile call.
And an average never tells you which twenty words it got wrong. The easy stretches come back perfect. The errors gather in the crosstalk, the accent, and the one term the matter turns on.
The refinery
Three stages. Nothing skipped.
A refinery does not improve its raw material. It separates what belongs from what does not, and it does that in stages, because one pass never gets it all.
02 / SEPARATION
Separation
The easy passages and the hard ones are pulled apart. Where a machine draft is used it is treated as raw stock, never as a finished document. Speakers labelled, timestamps set where you asked for them.
03 / REFINED
Refined
A person works the hard parts against the audio, replaying until the wording is certain. What will not come clear is marked with its timestamp, so you can see exactly where to check.
- +Speaker names only where confirmed
- +Timestamps at agreed intervals
- +Clean or full verbatim
- +Uncertainty marked with its timestamp
- +Editable Word document
- +South African English by default
Try it
Hold to listen again.
This is the moment a machine will not give you. Press and hold, the way a person plays a line back until it is certain.
…and the amounts in dispute are set out at paragraph twelve of his affidavit.
Press and hold the dot. A machine heard this once and moved on.
It invented a person. It flagged nothing.
What you get back
How closely should the page follow the recording?
It depends on the work. A research interview may need every hesitation left in. A board record usually reads better without them. The mode is agreed before anything begins, so the level of refinement is never a surprise.
- Clean verbatim Default
- Also called intelligent verbatim. The speaker's meaning intact, made easier to read. Repeated false starts and incidental fillers come out where removing them changes nothing that matters. Meaningful pauses, laughter, interruptions and colloquial phrasing stay in.
- Full verbatim
- Also called strict verbatim. It records speech as spoken, keeping repetitions, false starts, fillers, stutters, significant pauses and overlapping speech, with short bracketed markers where a non-verbal event affects the meaning.
- Edited transcript
- Grammar and presentation improved, still traceable to the recording. Labelled as an edited transcript so the level of refinement is on the record. Facts, commitments and the speaker's position are never changed.
- Research interview
- A deliberately conservative approach for qualitative work. Participant language and meaningful interaction markers are preserved. Names are used only where authorised or unambiguously confirmed, otherwise neutral labels or agreed pseudonyms.
Unless you ask for something else, the default is clean verbatim in South African English, with neutral speaker labels where names have not been confirmed. Transcription in English and Afrikaans, including recordings that move between the two. Where a speaker reaches for a word from another language, it stays in the transcript as it was said.
See the difference
The same interview, in both modes.
One recording, transcribed two ways. Switch between them and watch what full verbatim keeps and clean verbatim takes out.
Researcher00:00:04 – 00:00:15
Thank you for speaking with me, Maya. I just want to understand how you adjusted to postgraduate study in Cape Town. What was the first term like for you?
Um, thank you for speaking with me, Maya. I just want to, to understand how you adjusted to postgraduate study in Cape Town. What was that, the first term like for you?
Maya00:00:16 – 00:00:40
At the beginning, I felt overwhelmed. I travelled from Khayelitsha to Rondebosch three evenings a week, so getting to class on time was difficult. But after about a month, I made a study schedule around my work shifts and the commute became more manageable.
At the beginning, um, I felt, I felt overwhelmed. I travelled from Khayelitsha to Rondebosch three evenings a week, so, so getting to class on time was difficult. But after about a month, I made a study schedule around my work shifts and, and the commute became more manageable.
Researcher00:00:41 – 00:00:48
Where did you find support when the workload felt demanding?
Okay. And where, where did you find support when the workload felt, felt demanding?
Maya00:00:49 – 00:01:16
A classmate and I met at the library in Mowbray on Saturday mornings. We reviewed the readings before the seminar. That made a big difference because I could ask questions before I felt behind.
A classmate and I, uh, met at the library in Mowbray on Saturday mornings. We reviewed the readings before the seminar. That made a big difference because I could ask questions before I, I felt behind.
Researcher00:01:17 – 00:01:23
Was there anything in the orientation recording that you found difficult to follow?
Right. Was there anything in the orientation recording that you found, uh, difficult to follow?
Maya00:01:24 – 00:01:49
One instruction about the workshop venue was not clear. I marked it as [inaudible 00:01:31] and checked the detail with the course administrator instead of guessing. The place name sounded like Mowbray, but I was not certain, so I would write [unclear: “Mowbray”?] until it was confirmed.
One instruction about the workshop venue was, was not clear. I marked it as [inaudible 00:01:31] and checked the detail with the course administrator instead of, you know, guessing. The place name sounded like, um, Mowbray, but I was not certain. So I would write [unclear: “Mowbray”?] until it was confirmed.
Researcher00:01:50 – 00:01:56
What helped you feel more confident by the end of the term?
And what, what helped you feel more confident by the end of the term?
Maya00:01:57 – 00:02:17
I started preparing one question before each class. By the end of the term, I was contributing in discussions and helping another student find the right support services on campus.
I started preparing one question before each class. By the end of the term, I was, I was contributing in discussions and helping another student find the right support services on campus.
The two marked passages appear in both versions, and they are the point. Where a word would not come clear it is flagged with its position in the recording, so you can go to it and decide for yourself. A machine would have filled both of those in and told you nothing.
Rates and turnaround
Pick the pace the work needs.
| Category | Per minute | Delivered |
|---|---|---|
| Standard | R14.50 | 3 to 5 days |
| Express | R16.50 | 48 hours |
| Urgent | R18.50 | 24 hours |
Rates are per recorded minute and cover clear English audio, up to two speakers, intelligent or clean verbatim, and a standard document format. Express and Urgent delivery are confirmed against capacity before they are agreed. The minimum order is 15 minutes. All prices are inclusive, so what you see is what you pay.
More speakers, full verbatim, timecoding and difficult audio each change the figure. Every one of those rules is published in full, with the amount it adds, on the pricing page. Recordings with six or more speakers, and complex audio, are reviewed before a final quote is confirmed.
Confidentiality
Your recording is not training data.
Recordings, transcripts, names and contact details are treated as confidential client material. Nothing is quoted, reused, published or shown as an example without your written permission.
Personal information in a recording falls under South Africa's Protection of Personal Information Act, and it is handled on that basis. Where material is sensitive, the handling and the transfer method are agreed before you send anything at all.
Please do not attach a recording to the form on this site. Describe the project first and we will agree a secure way to get the file across.
Questions
The things people actually ask.
Why pay for this when AI transcription is nearly free?
Because free is priced right for audio that is easy. On a clean solo recording the machine is genuinely good and you should use it. Send us the other kind. The hearing with four people and a bad speakerphone is where the machine quietly invents things, and where a wrong word costs more than the whole transcript.
Can you guarantee a transcript with no errors?
No responsible transcription service should make that promise, and we will not. What we do instead is work to a structured review and mark material uncertainty rather than resolve it by guessing. How much review a recording needs depends on its audio quality, subject matter, number of speakers, and the scope we agree.
What happens when a word genuinely cannot be made out?
It stays visible as uncertain, with its position in the recording, so you can go straight to it. A passage that cannot be heard is marked [inaudible mm:ss]. A word that is probable but not certain is marked [unclear: word? mm:ss]. Where a speaker is not identified, neutral labels are used rather than a guess at a name.
What is the difference between clean verbatim and full verbatim?
Clean verbatim, also called intelligent verbatim, keeps the speaker's intended meaning while taking out verbal clutter that carries no substance. Full verbatim, also called strict verbatim, records speech as spoken, keeping repetitions, false starts, fillers, stutters, significant pauses and relevant interaction markers. Choose full verbatim where the way something was said matters as much as what was said.
Which languages do you work in?
English and Afrikaans, including recordings that move between the two inside a single conversation. Where a speaker reaches for a word from another language, as happens constantly in South African speech, that word stays in the transcript as it was said. It is not translated and it is not quietly dropped. What we do not take on is a recording conducted in a language other than English or Afrikaans. Translation is separate work and is never assumed. If you need it, tell us the source language, the language you want the record in, and whether the original wording should stay visible alongside.
How do I send you a confidential recording?
Not through this website. Use the form to describe the project without attaching anything, and we will agree an appropriate way to transfer the file before you send it. For sensitive material we also agree who is authorised to give instructions and how names and identifiers should be treated.
Can you remove names or identifying details?
Only when it is expressly asked for and scoped, never silently. It is also worth knowing that removing a name is often not enough on its own, because an employer, a location, a role or an event can still identify someone. We can flag the details that look identifying, but you decide what comes out.
What format does it arrive in?
An editable Word document unless you ask for something else. The format is agreed as part of the scope, along with a short delivery note listing what was supplied, the transcript mode used, and any limitations or assumptions worth knowing about.
The offer
Send us 5 minutes.
Pick the worst 5 minutes you have. The bad room, the heavy accent, the three way argument. We transcribe it and send it back, free, with nothing to sign. Tell us about it below and we will agree a safe way for you to get the file across.
Thank you. Your enquiry has been received.
We will review what you have sent and reply about the next step. Please do not send any recording until we have agreed how to transfer it.
Or reach us directly