shuka-v1

About Model

Shuka v1 is an innovative audio understanding model for Indic languages, combining Saaras v1 encoder and Meta's Llama3-8B-Instruct as the decoder. Trained on less than 100 hours of data, it outperforms larger models in audio-based question-answering tasks and supports fine-tuning for customized use cases. Shuka v1 is available open-source, marking the start of advancements in audio language models for Indic languages.