Typing out text from videos, images, or thumbnails on websites is often a tedious and error-prone task. This issue becomes particularly evident on platforms like YouTube, where valuable information is frequently presented in video content or thumbnail images. Users face challenges when they need to manually extract this text, especially if it includes links, technical terms, code examples, or mathematical equations.

Scenarios where this problem arises:

• Manually Typing Displayed Links: Typing out links shown in videos or thumbnails is inefficient and prone to errors since they cannot be copied directly.
• Copying References in Presentations: Extracting references or citations displayed in the footer of presentations is difficult and time-consuming.
• Copying Code to Editors: Transcribing code snippets from videos or images into a text editor is not feasible directly and is error-prone.
• Extracting Text from Images: Capturing text from documents shared as images on social media or other platforms is challenging.
• Feeding Text to LLMs: Users may need to extract text to input into language models for summarization or further processing.
  Addressing these challenges would significantly improve efficiency and accuracy in text extraction from multimedia sources.