Skip to content
PDFBetter
Upload
AI toolsBeta

Ask questions about an image

Upload a JPG, PNG or WebP image (HEIC in Safari) and ask questions in plain language; answers come from what the image itself shows.

Drop an image here

or paste with Ctrl/⌘+V

Uses AI (OpenAI, not used for training). 5 free AI credits without an account, 20 a month on the free plan.

JPG · PNG · WebP · HEIC · up to 25 MB

Try it 5× without an account. A free account includes 20 AI credits a month. See plans

What it does

Chat with an image, explained

Chat with an image answers questions about a picture: a screenshot of an error message, a photo of a receipt or whiteboard, a chart from a report, a scanned letter. Ask “What is the total on this receipt?”, “Which month had the highest value in this chart?” or “Transcribe the handwritten note”, and the answer is based on what the image shows.

Read more

PDFBetter turns the image into a one-page PDF in your browser and saves it to your documents, then sends the image itself to a vision-capable AI model with every question, so the model looks at the picture rather than at a text guess. Answers cite the image as page 1. Where the page has a text layer, quoted words are checked against it; text read from the picture itself can't be checked that way, so those answers cite the page without a quote.

When something is too small, blurry or cut off to read, the answer says so instead of guessing. For a multi-page scan, run OCR PDF and use Chat with PDF, which quotes and verifies exact passages.

Each question costs 1 AI credit. The image and your question are sent to OpenAI's API under OpenAI's API data-usage terms, which by default exclude API data from model training. Answers can be wrong and are not medical, legal or financial advice.

How to do it

Ask questions about an image in 4 steps

  1. 1Your PDF

  2. 2Ask or summarize

  3. 3Answers that cite pages

  1. 1

    Upload the image

    JPG, PNG or WebP: a photo, screenshot, chart or scan. iPhone HEIC photos work in Safari; in other browsers, export them as JPG first.

  2. 2

    Let it prepare

    The image becomes a one-page PDF in your browser and is saved to your documents. The original picture isn't changed.

  3. 3

    Ask a question

    Type your own or pick an example. Each message uses 1 AI credit. Without an account you have 5 AI credits to try it.

  4. 4

    Check and follow up

    Compare the answer with the image next to the chat, and ask follow-ups in the same conversation.

Example

Before and after

Before

A phone photo of a restaurant receipt and the question “How much was the tip, and what was the tax rate?”

After

An answer giving the tip and tax lines as printed, citing the image, with a note that the bottom edge of the receipt is cut off in the photo.

Settings & formats

What you can control

Input

JPGPNGWebPHEIC (Safari)

Output

Answers that cite the image (page 1)

Good to know before you start

  • The model can misread small, blurry, rotated or handwritten text; check numbers against the image.
  • Answers read from the picture cite the image (page 1) but carry no verified quote, because an image has no text layer to check the words against.
  • One image per conversation. For several photos, combine them with JPG to PDF, run OCR PDF and use Chat with PDF.
  • HEIC photos open only in browsers that can decode them (Safari); elsewhere, export the photo as JPG first.
  • It describes what is visible. It can't tell whether an image is authentic or edited.
  • In private AI mode, images are only sent to a model when your deployment's private endpoint supports images.
If something goes wrong
It says it can't read part of the image
Upload a sharper or larger version, or crop the image to the part you care about and ask again.
My HEIC photo won't open
Your browser can't decode HEIC. Open the page in Safari, or export the photo as JPG from your Photos app.

Chat with an image: questions people ask

Does the AI really look at the image?

Yes. With every question the picture itself, scaled to at most 2048 pixels on its longest side, is sent to a vision-capable model together with any text the page has. It isn't answering from a text extraction alone.

How are answers cited?

Answers point to the image as page 1. When the PDF page has a text layer, quotes are checked word for word against it; what the model read from the picture itself is cited at page level, without a quote, because there is no text to check it against.

What kinds of images work?

Photos, screenshots, charts, diagrams, receipts, handwritten notes and scanned pages in JPG, PNG or WebP, and HEIC photos in Safari. Sharp, upright images with legible text give the best answers.

How much does it cost?

1 AI credit per question. The Free plan includes 20 AI credits a month and Pro includes 500.

Is the image stored?

It is saved to your documents as a one-page PDF so you can come back to it, and you can delete it any time. Guest drafts are deleted 1 hour after they are created unless you create a free account.

Is my image used to train AI?

The image is processed by OpenAI's API under OpenAI's API data-usage terms, which by default exclude API data from training. PDFBetter does not train models on your files.

Upload a PDF