Tips for intelligent-verbatim transcribing

  • An intelligent-verbatim transcript should be theoretically capable of standing alone, without its accompanying video, as a complete record of a meeting. In archival terms, we create them against the eventuality that the video is lost, or its file format becomes obsolete.

  • Be aware that a transcript of a 2-hour meeting is around 18.000 words and can take 2 days to prepare. Take this into account when planning and when costing.

  • Verbatim transcription is a labour-intensive method. Only use it when accountability and transparency is needed long-term - for example, executive meetings where reps make decisions that affect others.

  • Your aim is to produce something readable, so good punctuation is helpful. If working from an AI text, remember that AI tools are not good at punctuation, so don’t hesitate to correct.

  • However, don’t attempt to correct a speaker’s grammar. The way someone phrases things is part of the content of their speech; so your approach needs to be descriptive rather than prescriptive, and it isn’t your role to “correct” the speaker’s idiolect to your local version of standard English.

  • Don’t go by what you think you heard, especially for references you’re unfamiliar with, names of people or organisations, etc. Google it, or contact the speaker and ask.

  • Usually, you’ll remove “ums and ahs” - but occasionally you might choose to leave them in if they reveal the speaker’s uncertainty or hesitancy and you feel this is important information.

  • Sometimes you might need to annotate. Any intervention you make to the text should be marked as such. Some useful annotations are:

    • [pause] if there’s a silence. Without it, listeners may think there is a fault with the recording.

    • [laughter] (if everyone does) and [laugh] (if only the speaker does).

    • [unclear] If you cannot tell what is being said despite your best efforts, don’t guess.

    • [overtalking] - when someone talks over another speaker, it’s best to note it as the reason why both speakers may be unclear. It can also give valuable clues to the mood of the meeting.

    • [misspeaking], if it’s clear that someone has done so - for example if someone says “September” when you feel it’s clear from context that they mean October, use [misspeaking - note of what you think was intended], rather than correcting it without comment, because there is always a chance that you’re wrong.

Last updated