Methods and workflow
Local transcription
Transcription that runs on your own machine, without the audio being sent to a server.
Local transcription means the model runs on your own machine. The audio is not uploaded anywhere, and the text is produced where the file already is.
The counterpart is cloud transcription.
What it solves
The privacy question, in its entirety. When the audio never leaves the machine, the whole row of questions that otherwise has to be answered disappears: who is the processor, where is the data stored, for how long, is it used for training, is it inside the EEA.
That is why running locally is the natural choice for:
- patient and client conversations
- personnel cases and disciplinary meetings
- research data containing personal information
- anything under a duty of confidentiality
It also works without internet, and there is no cost per minute.
What it costs
Hardware. The model has to run somewhere. On modern Apple Silicon it is comfortable. On an older laptop it takes substantially longer than in the cloud.
Speed. Batch transcription locally is usually slower than in the cloud. Real-time locally is more demanding still, since every second of audio must be processed in under a second.
Setup. Running an open model directly takes some technical work. Finished tools that do it for you exist, and are what most people should use.