This paper presents a Kannada-to-English speech-to-text system leveraging OpenAI’s Whisper model, fine-tuned for Kannada transcription. The proposed system addresses challenges in low-resource language processing by incorporating diverse datasets, preprocessing techniques, and a two-stage model pipeline. The first stage uses Whisper for Kannada speech transcription, followed by the MarianMT model for translation into English. Experimental results demonstrate significant improvements, with a Word Error Rate (WER) of 36.30% and a BLEU score of 0.69, validating the model’s performance. The novel approach shows strong resilience to dialectical variations and noise on a common web platform, facilitating cross-language communication in multilingual contexts making easy communication and interpretation of the language.