How to Ask an AI About a Screenshot or Photo
Screenshots, receipts and handwritten notes can go into a chat as images. Here is how to send one well, get an accurate reading, and what not to photograph.
Key points
- Deepent Research accepts JPEG, PNG and WebP up to 5MB each, three images per message, and image input needs a plan or credits.
- Shoot in good light, straight on, with the whole subject in frame, one purpose per image.
- Never photograph personal data, confidential work or faces, and verify any number the AI reads back against the original.
Contents
Retyping an error message character by character is tedious enough that many people give up and stop there. A manual is hard to copy from, a handwritten note cannot be typed at all, and the numbers in a chart have to be read one by one. An image avoids the whole problem: attach the screenshot or the photo, ask your question in a sentence, and the model reads what is in the picture and answers about it.
This guide covers how to do that well. It is written for people who have an image and want an answer, not a lesson in how the models work: what images can be read, the five habits that make the reading accurate, example questions by kind of image, and the photographs you should never send. The plain-text version of asking is in how to use a no-filter AI.
What can you do with an image in a chat?#
The model reads the text in the picture and then answers questions about it, so you can skip the retyping step. Usable images include error messages, receipts and invoices, charts and tables, handwritten notes, and foreign-language labels. The reading is not perfect, and small numbers are where mistakes happen most, so check any figure against the original.
Transcription means writing out the words contained in an image or a recording as text.
A few things people use it for:
- Error screens. Ask what the message means and what to check first, instead of searching the exact wording.
- Documents and statements. Have the items, dates and amounts pulled out, or just the text exported so you can copy it.
- Charts and tables. Ask which line grew fastest, or for the numbers behind a column.
- Handwritten notes. Have the content rewritten as a list, with unreadable words marked as such.
- Foreign text. Get the wording translated and the key terms explained.
How do you send an image and ask?#
Five steps, and the same ones on a phone or a desktop: prepare the image, open the service, attach it, write the question, then verify the answer. Attaching an image needs an account and a plan or credits, so it is not part of the logged-out text trial.
- Prepare the image. Take a screenshot for anything on a screen, or a photo for paper and objects. The habits in the next section are what make it readable.
- Open Deepent Research in a browser. Nothing is installed; a phone browser works.
- Attach the image. Deepent Research accepts JPEG, PNG and WebP files up to 5MB each, three per message, including photos taken with your phone camera. Image input requires signing in and a plan or credits.
- Write the question. One sentence naming what you want: "Explain this error and tell me what to check first." A stated purpose keeps the answer to the point.
- Verify the answer against the original. Dates and amounts especially: read them back from the image or the paper document before you act on them.
If you are still deciding which service to use, how to choose a no-filter AI compares them by the three things that matter, and the three ways to run one sets out hosted services, local models and APIs.
How do you get an accurate reading?#
Five habits decide most of the result: good light, a straight-on angle, the whole subject in frame, one purpose per image, and text that fills a reasonable share of the picture.
- Shoot in good light. In dim light the characters lose their edges, and the reading gets noisy.
- Shoot straight on. An angled photo distorts the text and makes it harder to read.
- Keep the whole subject in frame. A cropped edge can drop the row or the number you needed.
- One purpose per image. Several topics in one picture dilute the question, and the answer comes back general.
- Let the text fill the frame. Trim the empty space so the words occupy most of the image.
Open the image and read it yourself before sending it. If you cannot read a word, the model will struggle too, and you will spend the time again on a retry.
Screenshots should be sent at full size. Shrinking one to save bytes can make small text unreadable, which is the opposite of what you wanted.
What should you ask for each kind of image?#
Match the question to the image. The table gives a starting point for the five common cases; add the context you have, and the answer narrows further.
| Image | What to include | Example question |
|---|---|---|
| Error screen | The whole dialog | "Explain this error and what to check first." |
| Document or bill | The full page, no cropped edges | "List the items and amounts in a table, with the total." |
| Chart or table | Axes and legend included | "Which line grew the most? Include the numbers you read." |
| Handwritten note | The whole page, in good light | "Rewrite this as a list. Mark any word you cannot read." |
| Foreign text | The words sharp and in frame | "Translate this and explain the key terms." |
More uses, from research to creative writing, are in what people use uncensored AI for.
What should you never photograph?#
Do not send images that contain personal data, confidential work material, faces, or other people's contact details. Remember that a photograph carries everything that happened to be in the frame, not only the part you meant to share.
- Names, addresses, phone numbers, card numbers, order details.
- Faces of family, friends or colleagues.
- Unpublished company material, client documents, internal screens.
- Anyone else's contact details or records.
If a document belongs to your employer, check your organisation's rules before sending any part of it to an AI service.
On Deepent Research, input, including images, is not used to train AI, and the conversation in the logged-out trial is not saved. Other services differ, so read the privacy policy of whichever service you use before sending an image. Copy anything you want to keep, since the trial does not store it for you.
For anything medical or legal, such as a prescription label or a contract page, do not act on the reading alone. Confirm with a doctor, pharmacist or lawyer.
What is Deepent Research?#
Deepent Research is a browser-accessible LLM chat service built to refuse as little as possible. It keeps going with research that a standard AI would cut off mid-answer.
- It refuses nothing. We run uncensored and abliterated models with guardrails reduced as far as possible, so prompts a standard AI answers with "I can't help with that" still get a response (Qwen3.5 / Qwen3.8-27B-abliterated, depending on your plan).
- Your input is never used for training. What you look up is used only to provide the service, never to train AI or for any other purpose. Personal data is protected, and our privacy policy is public.
- Try it free. No GPU and no install, just open a browser. In the logged-out trial your conversation isn't saved, so you can start casually.
→ Try it free, no sign-up: https://deepentresearch.com
No question can stop research.
Frequently asked questions#
Can I ask an AI about an image?#
Yes. Attach a screenshot or a photo and ask your question in a sentence; the model reads the text and answers about the content. On Deepent Research, attaching an image needs an account and a plan or credits, so it sits outside the logged-out text trial.
What kinds of images work?#
Error screens, documents and statements, charts and tables, handwritten notes, and foreign-language text all read reasonably well when the image is sharp and complete. The failure cases are dim light, an angled shot, a cropped edge, and very small print: those lose detail the model needs.
How do I get a better reading?#
Shoot in good light, straight on, with the whole subject in frame, and let the text fill most of the picture. Keep one purpose per image, and send screenshots at full size rather than shrinking them. Then confirm any number against the original.
Is my image used for training?#
On Deepent Research, no: input including images is not used to train AI. Other services have their own policies, so read the privacy policy before sending anything sensitive, and leave personal data, confidential work and faces out of the picture entirely.
Disclaimer#
This article was drafted with AI assistance and then reviewed, fact-checked, and edited by the Deepent Research team before publication.
It is provided for educational and research purposes and is not intended to encourage unlawful activity or crime.
How this note was edited and verified
We cross-check primary sources such as official documentation and papers, and organize only what the facts support.
This note was drafted with AI, then fact-checked, expanded, and edited by the operator before publication.
Related notes
Writing With an Uncensored AI: Start a Story, Keep It Going
Tense scenes and technical drafts are where assistants stop. Here is how to write with an uncensored AI: brief it first, work scene by scene, and verify.
Why Does an AI Refuse? Four Reasons and What to Do Next
A refusal can come from four causes: a safety check, missing information, an unclear question, or a glitch or message limit. Read the message to find yours.
What Can You Hand to an AI, and What Should You Keep?
Summaries and background reading are easy to delegate. Personal data and publishing raw output are not. Here is where the line sits, and why, in five checks.
Try it yourself
Deepent Research is an LLM chat service that refuses almost nothing. Everything covered in this note can be tried without downloading anything.