🎶 Kaldi-ASR Speech to Text Scripts
Choose the first badge to view this repo on Binder with Kaldi-ASR all it's dependencies already setup.
I have chosen Binder
over something such as Colab
as the setup of the environment is to be done just once and on every usage request, it just reuses the same environment.
To view how the environemnt is set-up have a look at the Dockerfile.
If the Binder is under a "building" stage please wait for atleast 20 minutes. The Binder is updated for every push view a GitHub action trigger to be always up-to-date and maintained.
The speech_to_text.sh
script follows through the entire decoding. For using the script on your local machine, just changing the KALDI_ROOT
env variable would
get you setup entirely. Please open an issue if there are any crashes or bugs so that it can be resolved as soon as possible.
The pre-trained weights are already loaded into this repo under the exp/*
and can be used as such (results can be found here).
The script must work out of the box for such weights trained with the Kaldi-ASR framework.