Whether it is an expert interview for a master's thesis, a journalistic conversation or market research: at some point the question arises of how to turn the audio recording into a clean text. There are four options to choose from – typing it yourself, free tools, AI transcription or a professional service. This guide shows the pros and cons of each method, the realistic time required and what you need to consider regarding the data protection of your interviewees.
How do I transcribe an interview myself?
If you transcribe yourself, you play the recording in stages and type along with the spoken text. To keep this from becoming a test of patience, most people use special software such as f4transkript or f5transkript. These programs slow down the audio, automatically jump back a few seconds via foot pedal or keyboard shortcut and insert timestamps. This saves you from constantly reaching for the mouse and makes the work much smoother than with an ordinary media player.
The honest catch is the time required: for one hour of interview audio, as an inexperienced transcriber you should expect five to eight hours of typing. With several speakers, technical terms or an unclear recording, it quickly becomes more. Transcribing yourself is therefore mainly worthwhile if you only need a few minutes or want to know the material so well that listening pays off anyway.
- Advantage: full control, no data leaves your hands, free of charge apart from the software.
- Disadvantage: very time-consuming, tiring, prone to errors with long recordings.
- Ideal for: short recordings, individual interviews, tight budgets with plenty of time.
Which software is suitable for transcription?
For manual transcription, f4transkript and f5transkript (for macOS) are the de facto standard in German-speaking countries, especially at universities. They offer speed adjustment, automatic rewinding and export for analysis programs such as MAXQDA. Free alternatives are the open-source player oTranscribe or simply a media player with adjustable speed – however, these lack convenience features such as automatic timestamps.
If you transcribe according to fixed rules, you should decide on the appropriate set of transcription rules beforehand. In qualitative social research, the simple transcription rules according to Dresing and Pehl are widespread because they are easy to read and quick to learn. For fine-grained linguistic analyses, the considerably more elaborate GAT 2 is used, which records stress, pauses and overlaps precisely. For most interviews, a simple set of rules is entirely sufficient.
Is free AI transcription good enough?
AI tools automatically convert speech into text and deliver a rough draft in minutes – tempting compared with hours at your desk. With clear sound, a single speaker and standard German, the results are often usable. But as soon as dialect, technical terms, several voices or background noise come into play, errors pile up: misheard words, missing punctuation, mixed-up speakers. This is followed by laborious post-editing, which quickly eats up the time saved.
A second point is often overlooked: data protection. Many free AI services upload your recording to servers, frequently outside the EU. However, your interviewees have usually not consented to processing on US cloud servers. With personal or sensitive content, this is a real problem – more on this below.
When is it worth using a professional for transcription?
A professional transcription service takes the entire job off your hands and delivers a finished, correct text according to the set of rules you require. This is worthwhile if your time is short, the recording is demanding or the result needs to be citable. The calculation is simple: the eight hours you would otherwise spend at your desk go into your actual analysis instead.
- Several speakers: correctly attributing speech in focus groups and group discussions is manual work.
- Specialist vocabulary: medical, legal or technical terms are spelled correctly.
- Academic citability: transcription according to a fixed set of rules for your thesis.
- Data protection: a confidential service provider instead of an anonymous cloud.
- Time pressure: you need the finished text, not yet another task on your to-do list.
Data protection: what do you need to consider with interviewees?
Interviews almost always contain personal data – even the voice and the name count. This means the GDPR applies. Your interviewees should consent in writing to the recording and further processing, and you should know where the data is processed. Do not upload sensitive recordings to free cloud tools without checking. When using a service provider, check where the data is processed and whether a data processing agreement (DPA) is possible – often a requirement, especially for research data.