Computer Transcription
Computers are getting much better at transcribing human speech, but are still not perfect.
- Computer software usually does better with slow, enunciated speech without an accent.
- If you have an accent but your participants do not, you might try playing the original audio and see how it works.
- If you do not have an accent, it may be easiest to listen to the audio on headphones and speak it aloud to an automated transcription app, which also lets you add punctuation and speaker names and modify your voice to improve accuracy.
- All automated transcription requires having data on the internet. If your audio has any personal or sensitive information, you must be especially careful using these tools.
- The paid NVivo Transcription service advertises HIPPA compliance.
- If you re-speak the audio, you may be able to avoid personal information by substituting it with aliases or other keywords.
- You may be able to use automated transcription for parts of the audio and hand type other parts.
All Options
Free Online Transcription
Free Offline
For researchers who have concerns about uploading their data to an internet-based server, there are new AI-based tools that offer offline transcriptions that run locally on their computers.
| Pro |
No data is sent to the internet. |
| Con |
Require faster computers to achieve the best and quicker results, resulting in longer processing times on older and slower computers. |
- noScribe (Windows and macOS)
- Open-source
- Includes automatic speaker detection and understands about 99 languages (as of October, 2023)
- Offers an editor to manually correct the auto-transcripts
- aTrain (Windows)
- Includes automatic speaker detection and understands about 57 languages (as of October, 2023)
- Provides MAXQDA and ATLAS.ti compatible output
| University Libraries