AI-powered speech to text transcription

Automated transcription is a machine first pass that is then corrected by a person against the audio, which costs less than a transcript typed from scratch. The work is produced by people and not by software, and what shapes it here is audio, video and document material in almost any format, including legacy media, so the deliverable fits how you actually work instead of a generic template. Every file is checked by a second reviewer before it is sent.

Raw software output is never delivered as the finished file; a transcriber corrects it line by line against the recording. None of that is left to chance on the day: it is confirmed at quoting stage. Anything on physical media is digitised first, logged, and then worked from the digital copy.

It depends on length and on how difficult the source is, and the time is in the quote before you commit. Rush delivery within 2 to 24 hours covers the urgent end. Automated transcription is scheduled against the people who actually do this work, not against a general pool.

The standard is 99+%, reached by doing the job twice: one person produces the file, a second checks it against the source. What gets checked on automated transcription is the detail a reader would otherwise have to verify themselves: names, figures, terms and anything the source left ambiguous. Degraded tape and low-bitrate audio go to people used to that loss, not to software that fills the gaps with guesses.

Only the people doing the work can open your material, and all of them are under NDA. Archive material is often the only copy in existence, so originals are logged, restricted and returned. Automated transcription often involves material people would not want circulated, so the file is closed to everyone except the pair working on it. Transfer is encrypted in both directions, and deletion on an agreed schedule can be set up where your records require it.

Word, PDF, Excel, Google Docs or plain text, or your own template applied so the file comes back ready to use. Headings, tables and numbering are built properly instead of pasted in. The format for automated transcription is agreed at order stage, so the first delivery is the one you can file, send on or publish. Source material in almost any audio, video or document format is accepted, including formats most suppliers turn away.

Three steps: send the material, approve the quote, receive the finished work by email. Nothing starts before you approve the price, and revisions are free for 30 days after delivery. Say what the automated transcription is for when you send it, because a filing, a publication and an internal record are formatted differently. Say what the material is stored on, because an older format changes what has to happen before anyone starts work.

Send an existing machine transcript and it is corrected against your audio rather than retyped from the beginning. That is part of what automated transcription means here, not an extra, and anything genuinely outside the brief is quoted before it starts. A digital master of the source can be returned alongside the finished file where the original is fragile.

Yes. Automated transcription is available in 100+ languages, with every file going to a native speaker of the target language and a reviewer who also reads it. Older recordings often mix languages, and each one is kept as it was spoken instead of flattened into one. Where a language is rare, availability is checked before the quote.

General transcription starts at $1 per audio minute and legal transcription at $2, with volume pricing once a project passes an agreed size. The quote for automated transcription is worked out from how long the recording runs, how clear the audio is and how many people are speaking. Automated transcription is quoted from the material you send, so a short job and a long one are worked out the same way. The source format makes no difference to the rate; digitising legacy media appears as its own line in the quote. Send the material and the figure comes back within minutes.