Image to Text: Copy Text from Photos and Screenshots
Chat shortcut: /imagetotext
Image to text with AI text recognition: upload a photo or screenshot with your account and get the text. The model reads print, tables and usually handwriting too, and is meant to mark uncertain spots with [?].
With Account
Text recognition with an AI model
The tool hands your image to an AI model that is meant to reproduce the text word for word: headings, paragraphs and tables in the order of the image, handwriting deciphered as well as possible. It is meant to mark illegible spots with [?] rather than make anything up. As with any model reply, errors are possible, so double-check names and numbers.
You upload a photo, a screenshot or a scan and choose whether you want the text as in the image or as Markdown with tables and lists. In the task field you can say right away what should happen next: list only the amounts, translate into Spanish, correct the spelling or write a summary.
For a good result, photograph the document straight on, in good light and without shadows, so that the page fills the frame. Screenshots work well because the type is sharp and straight.
How it works
- 1
Upload an image
Sign in and drop a photo, screenshot or scan as JPG, PNG, WebP or GIF up to 30 MB.
- 2
Choose the format
"Text as in the image" for plain text or "Markdown" for tables, lists and headings.
- 3
Add a task (optional)
For example "only the addresses", "translate into Spanish" or "as a list".
- 4
Use the text
Click "Start", check the recognized text, copy it or carry on working in the chat.
Typical uses
Invoices and receipts
Read amounts, dates and line items from photos and reuse them as a table.
Lecture slides and whiteboards
Turn photos from a lecture into notes and have them summarized right away.
Screenshots of chats and websites
Copy text where selecting is not possible, such as from images, PDF screenshots or apps.
Handwritten notes
Photograph a notebook or whiteboard and save it as editable text.
Good to know
- Input: JPG, PNG, WebP or GIF up to 30 MB, upload with an account; one image per run; for the next one, keep going in the same chat. Images are reduced to at most 2,048 pixels on the longest edge before recognition; GIF is read as a still.
- Two outputs: text as in the image, or Markdown with tables, lists and headings; the model is meant to output uncertain spots as [?].
- Suitable for print, screenshots, tables and clearly legible handwriting; cursive, small type and poor photos raise the error rate.
- No language selection is needed; check accented characters, names and numbers in the result.
- Result and image stay in the chat history until you delete the chat; uploads without a chat are deleted after two days. Recognition runs through the selected AI model.
- Billed per reply by the model's actual usage (input and output, with thinking models also the thinking phase); no subscription needed.
Frequently asked questions
Which images work?
JPG, PNG, WebP or GIF up to 30 MB: screenshots, photos of documents, whiteboards, invoices, business cards and notes. The sharper, brighter and straighter the image, the better the result; very small type is best photographed in sections.
Does the AI recognize handwriting?
Clean block letters usually well, cursive with more errors. The model is meant to mark uncertain spots with [?]. Always double-check names, columns of numbers and codes.
Are tables and lists preserved?
Choose "Markdown (tables, lists)" and the model outputs tables as Markdown tables and enumerations as lists; check columns and numbers. "Text as in the image" gives you the plain text in the order of the image.
Which languages are recognized?
The AI models read the common languages without a language setting; check accented characters and special characters in the result. You can also request a translation right away in the task field.
How accurate is the recognition?
With clear print usually close to the original; with small type, poor light, shadows or an angled shot the error rate rises. For longer documents, test one page first.
Can I translate or summarize the text afterwards?
Yes, the text ends up in the chat, where you can translate it, shorten it, have errors corrected or ask questions about the content.
What does it cost and what happens to my image?
Billed per reply by the actual usage of the selected model (input and output, with thinking models the thinking phase); no subscription is needed, the upload requires an account. Image and text stay in the chat history until you delete the chat; uploads without a chat are deleted after two days. Recognition is done by the selected AI model under the provider's terms.