Xhosa transcription services

It is a language of respect some Xhosa women observe, avoiding their father-in-law's name and its syllables by substituting other words. Researchers have found that some of these substitutions are invented by an individual woman for her own household, so they exist nowhere in any dictionary. A transcriber who does not recognise the custom risks treating a genuine, deliberate word as a mistake or a name to chase down.

By recognising what it is doing rather than looking it up. If context, generation and family relationship point to a respect substitution rather than an error, our transcriber notes it as such instead of flagging the audio as unclear or guessing at a homophone. This is a case where linguistic and cultural knowledge does work a glossary cannot.

It has genuinely declined and is more alive in rural areas than in cities. Researchers working in Cape Town found most women they spoke to no longer used the vocabulary regularly and were unsure whether the custom was still practised at all, with English and Afrikaans increasingly filling the gap. We do not assume it is present in every recording — we listen for it, and expect it more in some settings than others.

Because a speech-to-text model can only match sounds against a fixed vocabulary, and a household-specific respect word may exist nowhere else, ever, for any model to have learned. It will either mistranscribe the sound as the nearest real word it knows or flag it as unrecognised. A native Xhosa transcriber who understands the social context is what tells the two apart.

There is no flat rate, because no two files take the same work. Runtime, recording quality, how many people are talking, your deadline and whether you also need an English version all move the figure. Every quote carries our Price Match Promise.

Upload is encrypted, the file goes to one named Xhosa transcriber who has signed an NDA, and nothing is put through public speech-to-text services at any point. If you have your own confidentiality agreement we sign it, if you have retention rules we work to them, and if your policy says the audio goes once the transcript is approved, we delete it.

It is normal work. Phone audio, site noise, focus groups and meetings where three people speak at once all get transcribed. What difficult audio changes is time, and occasionally certainty about a word or a name. We listen to a sample first and tell you straight what we expect to be able to capture, rather than handing you a file peppered with inaudible tags.

In whatever the next step needs: plain text, Word, PDF, SRT, VTT or your own template, with timecodes and speaker labels set the way you want them. Click consonants are delivered in standard Unicode Latin script so they display correctly wherever the file lands. Tell us what happens to the transcript after we send it and we format it for that, not for us.