kolodny/torah

Run the following:

pip install -r requirements.txt
python -m allosaurus.bin.download_model
python -m allosaurus.bin.update_phone --lang heb --input heb_phones.txt

You also need to have node installed and run the following to install the dependencies:

npm install

You then need to run these two commands to start the server:

npx tsx --watch server.ts

npx run dev

Now go to http://localhost:5173/ and you should see the app running.

The basic flow of the app is as follows:

• User goes to the app and chooses which text to read (right now we only have one text)
• User records themselves reading the text
• Audio is sent to the server which uses allosaurus to extract the sounds it heard
  • For each distinct sound, allosaurus will return a list of what it thinks the sound is
• The server sends the sounds back to the browser
• The browser uses a sequence aligner algorithm to try to find the best match between the sounds and the text
  • The algorithm goes through each distinct sound heard and sees if any of the candidates match the expected sound
• The browser displays the results in a diff-like format