Advice on develop an offline audio transcription software with no knowledge

[woah]

Hi all, my job requires me to do lots of audio typing. These audios involve lots of medical description. Due to confidentiality, I can't use those online transcription softwares such as otter.ai. So, I am think to build a offline audio transcription software by myself to increase my work efficiency.

However, I do not have any knowledge regarding coding and programming. I am hoping someone can give me direction on how to develop such software, and whether it is possible to do it. For example, give me some keywords, so that I can search it by myself. I got some audio files (with different speakers) and the transcriptions on hand. I am thinking it is possible to teach the computer to learn that AAA (sound) is equal to XXXXXX (word), so that the computer can generate the corresponding text next time when I input another audio file but with another speakers. And, for the whole process, I don't need to connect to the cloud or something, so that no one apart from me can access the files (due to security and confidentiality).

I tried to do some searching by myself. Some people talk about API, which I don't really understand. My concern is still about the confidentiality. To me, it seems like the company that providing the API can access my data, which I don't want to.

Thank you for your advices in advance.

[DeaD_EyE]

You should try this project: https://github.com/openai/whisper
I've used it to transcribe some Videos, and it worked well with German.
It works even better with english. It is also possible, that you create your own models.

For offline translation, you could use: https://github.com/argosopentech/argos-translate

[jefsummers]

This would be a big project for learning to program. Working as suggested above, or consider buying an existing product. I have personally used Dragon voice recognition software (medical edition) for dictating into EHRs, as well as M*Modal dictating into EPIC.

Here is a set of reviews. Commercial products have solved the HIPAA question or they would not exist.

[Larz60+]

As well as above two suggestions, you can search here for other modules similar to whisper, should you choose to try writing the code it yourself.

This is an easy place to search packages, and usually contains the contents of the GitHub repository Readme.md file.

In the instance of whisper, it will direct you to 'Graphite' which is a package that contains three components, which all work together.
One of those three components Graphite contains is 'whisper'.
The GitHub URL for all components, can be found under 'Active Development' title on that page.

There are, as of today, 426,423 prewritten packages in pypi, that can be used by your software (including a page about whisper).

Ultimately, all will drill down to the GitHub page (provided by DeaD_EyE above), where the source code can be found (or get a quick and dirty download by clicking the 'Download files' button on pypi.