🔍 Lupa Inteligente

How to scan a product label and extract readable text

Labels are difficult documents for optical character recognition (OCR). A bottle may curve away from the camera, shiny packaging can create reflections, and ingredient lists are often printed in tiny lettering. This browser-based OCR scanner is useful for turning visible text into something you can review, copy and reuse. It should be treated as an assistant for transcription, not as an independent way to verify whether a product complies with applicable labeling rules.

Take a better picture before running OCR

Place the package under diffuse light so the printed surface is evenly illuminated. Avoid direct reflections, especially on glossy plastic and metallic labels. Keep the camera as perpendicular as possible to the text, stabilize your phone and fill the frame with the relevant part of the label. If the print wraps around a bottle, photograph it in several sections rather than forcing all the text into one distorted image. A clear close-up usually improves recognition more than repeated attempts on a blurry photograph.

What kinds of labels can you digitize?

You may use the tool for product descriptions, ingredient lists, storage guidance, serial information and text printed on packaging. It can also help transcribe small instructions for cataloging or accessibility tasks. Results depend on the language, typography, contrast and image quality. Unusual fonts, handwriting, damaged labels and curved surfaces may produce missing characters. Text extracted from a label is not proof that the printed statement is correct.

Reviewing OCR output before reusing it

• Numbers and codes: compare batch numbers, expiration dates and measurements character by character.
• Ingredients: check punctuation, hyphens and similar-looking letters in chemical or INCI names.
• Units: distinguish mg from g, mL from L and decimal separators where a transcription mistake could change meaning.
• Warnings: ensure negations and mandatory cautions are not omitted in the extracted text.
• Reading order: verify that text from separate columns or labels has not been combined into a confusing sentence.

What about translation?

OCR and translation solve two different problems. OCR tries to reproduce characters in the source language, while translation attempts to convey their meaning in another. If the original text was recognized incorrectly, its translation may also be wrong. Technical terms, regulated warnings and product directions can require specialized terminology. When precision matters, first correct the recognized source text and then verify the translation independently.

Example: documenting a cosmetic ingredient label

Suppose you photograph a small cosmetic container with a list of ingredients. Use good lighting, extract the text and compare every ingredient name with the package. Retain a reference to the original image in your own records. If you are building a product catalog, separate the product name, ingredients, net quantity and directions into their respective fields. Do not assume the OCR output automatically determines whether the labeling is compliant with health regulations; that requires a different review.

Privacy and technical boundaries

Image recognition is designed to run locally in your browser. This avoids uploading the selected picture to a dedicated transcription service, although the webpage itself can still use standard website resources. Large photographs may require significant memory, and some browsers will perform more slowly than others. Close unrelated tabs if you encounter device-memory limitations, or crop an oversized image before retrying. For image-size reduction, use our image compressor. For a Spanish interface, visit Escáner de etiquetas.

Safety note: never rely on unverified OCR or machine translation to determine medicine dosage, warnings, allergens or other safety-critical information. Read the original label and ask a qualified professional where appropriate.