Automatic Speech Recognition (ASR) is the task of converting spoken audio into written text. ASR focuses on “what was said” by producing the correct transcript. 

In this competition, you will build a model that performs Bangla long-form ASR on real long-form audio—lectures, interviews, conversations, and more. Given Bangla audio recordings, predict the corresponding Bangla text transcript. 

Goal: Produce an accurate Bangla transcript for each test audio file. 

Note: Participants do not need to handle punctuation explicitly. 

This competition is part of BUET CSE Fest 2026 and is hosted on Kaggle. 

Detailed Rulebook: https://tinyurl.com/DLSprint4Rulebook