Tunisian Speech Recognition

This is a Tunisian ASR based on the WavLM Model, fine-tuned on a dataset of 2.5 hours, resulting in a W.E.R of 24% and a C.E.R of 9%.

Interested? Try it out!

⚠️ Disclaimer:
This is a demo model, The transcription accuracy isn't accurate due to Hugging Face model storage constraints.
For better performance,you can run the full model locally.
Please check out the repository and follow the instructions: Full Model Repo Link