Freelance work · idea 49 of 52
Transcription for lawyers, researchers and podcasters
Automatic transcription is cheap and wrong in the places that matter. Professionals pay for the version they can quote, file or publish.
- Who pays you
- Law firms, academic researchers, journalists and podcast producers pay you per audio minute or per finished hour, by invoice, for verified transcripts, timestamps and speaker labels.
- Outcomes
- 0 receipts. Nobody's reported back on this one yet. Receipts open when accounts launch.
How it works
Speech-to-text software gets most words right and a lot of the important ones wrong: names, technical terms, crosstalk, mumbled numbers. Lawyers, researchers running interviews, and podcast producers publishing show notes need a transcript that's actually correct, and they pay per audio minute for a human to produce or verify one. You use the software for the first pass and your ears and knowledge for the second. Specialising (medical, legal, one academic field) is where rates go up.
What it actually takes
- Typing speed and a good pair of headphones. A foot pedal helps and costs little.
- Software: a transcription editor and a speech-to-text service for the first pass. Modest monthly cost.
- Vocabulary in one or two fields so you know that the word was "tortious" and not "tortoise".
- Confidentiality habits. Legal and research material is often sensitive; some clients require agreements.
Where it goes wrong
- Marketplace rates. General transcription sites pay a fraction of what direct professional clients pay. Go direct.
- Bad audio. Quote higher, or decline, for recordings made in a café with four speakers.
- Skipping the verification pass because the software looked fine. That's what the client is paying you not to do.
- Assuming the work will vanish. Volume of raw transcription is falling; verified, formatted, specialised transcripts still sell.
Published 28 Jul 2026