Choose a readable image
Upload a PNG, JPG, JPEG or WebP screenshot or photograph. Prefer a sharp, upright page with good contrast; if several images form one document, arrange the queue in the order you want the extracted text to appear.
Image to Markdown
Choose a screenshot or photo and extract its text into Markdown with local browser OCR.
Image OCRPrivate conversion workspaceNo account required
A clean document starts here.
Drag and drop your files, or choose them from your device.
.jpg · .jpeg · .png · .webp
10 MB per file · 5 MB textUp to 5 files per batch25 MB max total
Image OCR runs on your device with an English recognition model; your pictures are not sent to an OCR service.
How to use Image to Markdown
Proofread the recognized content
Read the live Markdown editor beside the visual preview, checking GFM table boundaries and the LLM token count. Correct uncertain letters, account numbers and currency values before relying on a recognized passage as an exact transcript.
Keep the text, control the image
Download the corrected Markdown or copy it into your notes. Recognition stays inside the browser, and the original image remains on your device; wipe optional recent-output history when you have finished with sensitive material.
Formatting fidelity, explained
Local OCR turns visible characters into editable text without requiring an external recognition account. It is useful for a receipt, a screen capture or a photographed handout when there is no selectable text to copy from the original material.
A clearer source reduces ambiguity before any formatting is applied. Crop unrelated surroundings, straighten a tilted photo and avoid glare across important figures; sharpening an unreadable image cannot reliably recreate a character that the camera never captured in the first place.
The resulting text passes through the Markdown workspace so you can form headings, lists and inspectable table rows. Formatting improves readability, but a properly aligned number is not proof that OCR recognized its digits or decimal separator correctly.
An ordered batch can combine photographed pages into one reviewable narrative or keep separate outputs in a ZIP. Give page images meaningful names and inspect the queue before conversion so a receipt’s back page or a report appendix does not interrupt the main sequence.
Image text recognition is most useful as a drafting aid with a clear verification step. The supplied English model works best on legible printed English, while handwriting, decorative fonts and mixed scripts can reduce accuracy. Compare visually similar characters such as zero and the letter O, then check punctuation in dates and identifiers. A screenshot of code warrants particular care because a missing quote or bracket can make an otherwise plausible-looking example unusable.
Once the words are editable, you can prepare a smaller context pack for an LLM instead of repeatedly uploading images elsewhere. Remove personal identifiers, add a short source label and use the token estimate to gauge the length of your transcript. Do not assume that an OCR table preserves every implied relationship in the picture: reintroduce clear column names and explain what an empty field means before asking another system to summarize or analyze its contents.
Frequently asked questions
Why does a sharp image improve Markdown output?
A sharp image gives the recognizer distinct character edges and spaces between words. Blur, compression, shadows and a slanted camera angle make several possible readings look alike, particularly in small tables. Retake or crop the picture before conversion where possible, then check important names and figures against the original. Enlarging a blurred image can make it bigger without adding the missing visual detail.
Can five photographed pages be put in order?
You can add multiple photos to the queue and arrange them before running the conversion, within the current file and batch limits. Choose a merged Markdown result for continuous pages or separate outputs when they represent independent documents. Confirm the sequence using filenames and previews, because a camera’s creation order may not match the report’s printed page numbers or the order you intended to read.
Does table alignment correct OCR number mistakes?
No, table formatting can repair column boundaries and improve alignment, but it cannot establish whether an image contained 8, 3 or another similar-looking digit. Inspect amounts, minus signs, percentages and totals directly against the picture. Where recognition is uncertain, correct the value yourself or mark it for review; a tidy table should never be used as a substitute for checking the source evidence.
Where does OCR run, and which language is supported?
OCR runs in the browser using the included English recognition model, with no need to send the image to an external recognition service. Printed English generally offers the clearest results; other scripts and complex handwriting are not guaranteed. The first recognition can take longer while the model starts, and large images use more device memory, so a desktop browser may be more comfortable for a substantial batch.