eddiegulay / swahili-asr-api

Javascript inference code to finetuned whispher model with mozilla common voice

Swahili-ASR-API

Python & Javascript inference code to finetuned whispher model with mozilla common voice

Note: The code in here work with finetuned model on swahili common voice dataset. The model is available at Whisperer_Mozilla_Sw_2000.

How to use

Just clone the repo and select the language you want to use. The code is written in python and javascript. The python code is used to run inference on the model using requests module and the javascript code is used to run the inference on the browser.

Javascript Web Inference

For direct browser inference check web folder. It contain simple HTML file index.html that will allow you to upload file and get transcription.

Javascript ~Node Inference

For node inference check src/index.js. It contain js request snippet that will allow you to specify file location, make RESTful request and get transcription.

Python Inference

For python inference check python/inference.py. It contain python code that will allow you to specify file location, make RESTful request and get transcription.

Important

Those Inference codes can be found in huggingface repository for the model and also you might want to get access token to use them. I'm totally cool to share the token with you. Just contact me on

• twitter
• linkedin