What Is Autrans? Nigerian Speech-to-Text, Explained

Most people meet Autrans with a specific problem in hand. They have a two-hour interview, a church service, a board meeting, or forty WhatsApp voice notes from a field team, and they need the words written down.

So the short answer: Autrans is a speech-to-text service. You upload an audio or video recording and it returns a written transcript you can read, search, edit and export.

The longer answer is about which recordings.

The problem with "supports English"

Plenty of transcription tools claim to handle English. Most of them mean one particular English — the one spoken in California or the Home Counties, in a quiet room, by one person at a time.

That is not what a Nigerian recording sounds like. A Lagos standup meeting runs in English until somebody makes a joke in Pidgin. A market research session in Ibadan moves between Yoruba and English depending on who is speaking and what they are describing. A sermon might open in English, quote scripture in Igbo, and land the point in Pidgin.

None of that is unusual here. It is just how people talk. But a system tuned on a narrower kind of speech treats it as noise, and you get a transcript full of confident nonsense.

Autrans exists because that gap is the normal case for a huge number of people, not an edge case.

What it actually handles

Nigerian English and Nigerian Pidgin, with partial support for Yoruba, Hausa and Igbo — including recordings where the speaker switches between languages mid-sentence. Arabic, French and Spanish are supported too. You can see the full list on the languages page.

"Partial" is doing real work in that sentence. We would rather name the limit than let you discover it in a transcript you were counting on. For a broader look at why these languages are hard for speech systems, there is a longer piece on why African language transcription matters.

What you get back

A transcript, labelled by speaker. That last part matters more than it sounds: an interview where you cannot tell the question from the answer is a wall of text, not a record. The mechanics of how that works are covered in speaker diarization.

From there you can search it, edit anything that came out wrong, translate it, generate notes and action items, or ask questions about what was said. Export it when you are done.

Who uses it

Two shapes, mostly.

Individuals with a recording and a deadline — journalists, researchers, students, podcasters, lawyers. They upload, wait, and take the transcript away.

And organisations: a university department, a newsroom, a research team, an NGO with people in the field. Several people share one pool of hours on a single invoice, which is usually the difference between a tool someone expenses quietly and one a department can actually adopt.

What it is not

It is not a live captioning service. Autrans reads a complete recording before committing to the words, which is a deliberate trade — full-sentence context is exactly what tonal languages and code-switching need. If you want the reasoning, real-time vs batch transcription covers it.

It is also not magic. Bad audio produces bad transcripts, in any language, with any tool. If your recording is muffled or noisy, fixing that first is the highest-leverage thing you can do, and there is a guide on cleaning up a WhatsApp voice note that applies more broadly than its title suggests.

Starting

A new account includes free transcription minutes each month with no card required, which is enough to run a real recording through and judge it on your own audio rather than on a page like this one. Plans and per-hour options are on the pricing section.

The honest test is your own worst recording. Sign up and try that one first.