“Live! Character Recognition” is comparable to optical character recognition software where input is fed through a camera attached to the computer and the processing is done at the instant of taking input. The goal is to recognize only the text component from the image stream which may have text-characters buried in the colored background, complex patterns and text like structures. The camera attached to the computer takes the live moving video or the image streams as an input then the text part of the image is only recognized through the use of a method, Multiscale Edge-Based Text Extraction. Finally, the use of Kohonen neural network helps to identify the text. This project is live i.e. all the processing is done to the image streams captured by the camera at that particular instant of capturing, hence no offline storage or buffering of large video is required. The programming platform that is used to develop this project is Java with the support from ImageJ for image processing and vlcj for capturing image streams.