Image-to-text extracts or generates text from images through OCR or visual captioning/description. Two meanings: OCR: Extract printed/handwritten text from documents, signs, screenshots (text literally in image). Captioning: Generate natural language descriptions of visual content (what the image shows). OCR technology: Deep learning OCR (Tesseract, EasyOCR, PaddleOCR), document AI (AWS Textract, Google Document AI), scene text recognition. Captioning models: BLIP, BLIP-2, LLaVA, GPT-4V, Gemini Vision - vision-language models generating descriptions. Dense captioning: Describe multiple regions of image in detail. Visual QA: Answer specific questions about image content. Document understanding: Extract structured information from forms, tables, invoices. Implementation: Vision encoder + language decoder, cross-attention or prefix tuning, trained on image-caption pairs. Use cases: Accessibility (alt-text), content moderation, visual search, document digitization, photo organization. Evaluation metrics: BLEU, CIDEr, SPICE for captioning. Challenges: Hallucination in descriptions, fine-grained details, counting accuracy. Foundation for multimodal AI applications.