Local transcription of large audio files

The solution automatically transcribes large collections of audio recordings on your computer. You add the recordings and start processing — the system recognizes speech with acceleration on an NVIDIA graphics card and saves a separate text for each file. It also suits lectures of more than two hours: there is no need to split them into parts.

After the initial setup, recognition works fully offline. Audio does not have to be sent to external services, and the volume of work is not limited by a subscription or API quotas. In one run, about 60 hours of lectures were processed in about 5 hours on a GPU.

For further work with the text there is an additional scenario: the finished transcript is passed to n8n, AI picks out the main points, structures the material and automatically formats it as an article in Google Docs.