Methods and workflow
Batch transcription
Transcribing a finished recording afterwards, rather than while the speaking happens.
Batch transcription is sending a finished audio recording for processing and getting the text back. The counterpart is real-time transcription, where the text arrives while the speaking happens.
Why it gives better results
The system has the whole recording available. It can use what was said later to interpret an unclear word earlier, and it can make several passes over the material.
It also gives better diarisation: grouping voices is easier when you have heard all of them through than when you have to decide as you go.
The difference shows most clearly on proper nouns, jargon and telling speakers apart.
Speed
The processing itself usually takes a few minutes per hour of audio in the cloud. Locally it depends on the hardware. Either way it happens without you having to follow along, which is the whole difference: machine time costs you nothing in attention.
When to choose batch
- Meeting summaries and documentation
- Interviews that will be quoted
- Research data
- Podcasts and video to be captioned
- Anything where the text will be used for something, not just followed along with