AI-powered YouTube video dubbing pipeline. Automatically transcribes (Whisper), translates (Google), and generates neural dubbing (Edge-TTS) with smart audio-video synchronization and background music preservation.

A high-performance Docker container that runs OpenAI's Whisper model. Optimized for CPU, Intel NPU, Intel Arc/iGPU, and NVIDIA CUDA GPUs.

A real-time Speech-to-Speech translation pipeline (ASR ➡️ NMT ➡️ TTS) using OpenAI Whisper, MarianMT, and gTTS. Features a Flask backend and a responsive web UI for low-latency multilingual communication and audio synthesis.

Dictator – Supercharge Cursor Chat with voice-to-text, custom AI prompts, and workflow automation. Speak your ideas, inject templates instantly, and code faster with AI-powered assistance.

Listening Between the Lines: An explainable multimodal framework for MCI detection from spontaneous speech. Leverages Selective State Space Models (Mamba) and Gated Fusion to integrate linguistic disfluencies and eGeMAPS biomarkers across multi-corpus benchmarks (Pitt, ADReSS, TAUKADIAL)

Analysis of Robot Ego-Noise impact on ASR models (Whisper) & Signal Processing solutions.

A real-time audio streaming POC featuring Voice Activity Detection (VAD), Faster-Whisper ASR, NLLB-200 translation, and Piper TTS. Built with FastAPI and React to demonstrate a low-latency, end-to-end speech-to-speech pipeline.

A specialized AI-powered educational tool for mastering mental arithmetic. Features local LLM integration (Llama 3), real-time voice transcription (Whisper), and an interactive Canvas-based Soroban abacus.