One to use as Automatic text detectionThe process of conversion is crucial to the eventual success of multimedia content retrieval. In this work, we offer an algorithm that can automatically recognise horizontally aligned text in photos and then isolate it from its surrounding context, regardless of how complicated the backdrop may be. The suggested method relies on the use of a color-reduction methodology, an edge recognition method, and the positioning of text sections via studies of projection profiles and geometrical features. As well we are using some files with the help of deep machine levelbased technique. And we are detecting text from the real time video also. The text is embedded in a wide variety of papers and natural settings, and it offers information that is both crucial and beneficial. It's not easy to pull text from a natural picture. We recognise and extract the text in a manner that makes it easily readable by a human. Here, we offer a quick approach to text localization by using an edge detection and clustering approach to pinpoint all potential picture edges.