The accessibility of open-source software is greatly aiding the popularity of speech recognition and deep learning. As an illustration, Kaldi has become a well-established framework for creating cutting-edge speech recognition systems in current times. This paper provides sum- mary of the various ASR systems based on Kaldi for Indian Languages. The results acquired using different models and classifiers on multiple datasets are strictly observed and compared. Experiments on various Indian datasets and objectives demonstrate that Kaldi can be success- fully utilized to create advanced speech recognizers. It offers a speech recognition system that utilizes finite-state transducers, accompanied by comprehensive documentation and scripts for constructing full-fledged recognition systems. The phonetic context modelling capability of Kaldi, which encoded in C++, is enabled by its core library. Along with stan- dard Gaussian Mixture Models, it also makes acoustic model-ling us- ing Subspace Gaussian Mixture Models (SGMM) easier. Additionally, a large variety of popular linear and affine transformations are supported, including support for a wide range of them.