This project focuses on providing speech recognition across industries and languages.
The project is being developed on a private basis with support from A\Ware and Reibke. Currently I'm focusing on languages used in the German area, but I'm happy about any support for other languages.
With the exception of the audio dataset and training script (at the moment, coming in future), all code, used models and datasets are released under open source license. An example of training large ASR models on small hardware take a look at the simplepeft project https://github.com/flozi00/simplepeft
Install the ffmpeg package for loading the audio files and python3 to run the code
With both packages installed you can just run "pip install -r requirements.txt" to install the python librarys and then start the server with gradio UI using "python app.py" For faster deployment you can even use the docker container https://hub.docker.com/r/flozi00/asrasaservice