Transcribing Audio and Video (SC)

Manually transcribing interviews can be time-consuming. At Swarthmore, the best automated transcription option for research is usually Adobe Premiere Pro. Students can download Adobe Creative Cloud to their personal laptops or use the ITS Media Lab computers.

How should you choose a transcription tool?

Consider the language(s) being spoken in the recording and whether these are supported by the transcription tool.

Consider the privacy and security needs of your research participants. For research, you may not have permission to share your recordings with a third-party service that requires uploading your conversations to its servers. Often Adobe Premiere Pro will be the best option here. OpenAI's Whisper can be run locally (i.e. on your computer, not online), but it's a little more complicated - learn more. 

Tips to make the process go smoothly:

• Start early. Plan to spend some time processing your recordings and troubleshooting the process.
• Automated transcription can be slow on some computers. Consider using the ITS Media Lab computers.
• Working with smaller files can help. This can involve using a lower-quality version of the file or breaking up a longer recording into segments.
• If you recorded a video, consider uploading an audio-only file to generate a transcript.

How can you use automated transcription effectively and ethically?

Automated tools don't handle all voices and terminology equally well. Even under perfect conditions, automated tools still regularly make mistakes.

Regardless of what transcription tool you use, you'll need to manually check and correct the output to make sure it actually reflects what your participants said. Sometimes mistakes will glaring and easy to notice, like when names are incorrect. But subtly misleading errors are also really common. Protect the integrity of your work by reviewing transcripts carefully against the original recording and correcting the transcripts as needed.