UnmarkPro
Guide and free tool

How to See the Prompt of an AI Image And why some images do not have one

Many AI image apps write the prompt and settings into the file itself. Drop the image below to read them. If nothing is there, the guide explains why.

Choose image

or drop PNG, JPEG or WebP files here. Add several at once.

Your images stay on your device.

The short answer

  1. Find the original file the AI app saved, not a copy from a chat or social site.
  2. Drop it into the tool above.
  3. Copy the prompt, or download the ComfyUI workflow if there is one.

Which apps save the prompt in the file

Where the image came fromPrompt in the file?
Stable Diffusion (AUTOMATIC1111, Forge, SD.Next)Yes: prompt, negative prompt, seed, sampler, model
ComfyUIYes: the prompts and the whole workflow
NovelAI, InvokeAI, Fooocus, SwarmUIUsually yes
Midjourney (downloaded file)Often: the prompt text and its settings
ChatGPT, Gemini, Adobe FireflyNo prompt. Only which app made it
Any image from Discord, X, Instagram, WhatsApp, or a screenshotUsually nothing left

Settings in each app can change this. For example, Stable Diffusion web UIs have an option to stop saving the prompt.

Why the prompt disappears

The prompt sits in a hidden part of the file, next to the picture. When a site re-saves an image to make it smaller, that hidden part is usually thrown away. That is why the same image shows its prompt on your computer but not after you post it and download it again.

Communities that share prompts, like Civitai, often keep the original files. If you want the prompt of an image you found online, look for a "download original" button, or ask the person who made it for the PNG.

Sharing your own images? Check what they give away

The same hidden data that helps you learn from other images also shares your prompt, seed and model when you post yours. If you would rather keep your prompt private, remove it before you share.

  • Drop your image into the tool above to see exactly what it reveals.
  • Use the AI metadata remover to save a copy without it.

What this cannot do

  • It cannot recover a prompt that was never saved or was removed. It only reads what is in the file.
  • ComfyUI graphs built with rare custom nodes may not show the prompt text, but the workflow download still works.
  • An image pasted from the clipboard usually has no prompt, because the browser re-saves it. Choose the file instead.

Related Guides & Topics

Frequently asked questions

How do I see the prompt used to make an AI image?

Drop the original file into the tool on this page. If the app that made it saved the prompt inside the file, you will see the prompt, the negative prompt and the settings, with a Copy button. Nothing is uploaded.

Why does it say there is no prompt in my image?

Most likely the file was re-saved somewhere. Discord, X, Instagram, WhatsApp and many other sites remove this information when you upload. Screenshots never have it. You need the original file the AI app saved.

Can I get the prompt from a Midjourney, ChatGPT or Gemini image?

Midjourney downloads can include the prompt text and its settings. ChatGPT, Gemini and Adobe Firefly save which app made the image, but not the prompt, so the tool shows the app name and says the prompt is not included.

How do I load a ComfyUI workflow from an image?

ComfyUI saves the whole workflow inside its PNG files. Drop the PNG into the tool and tap "Download workflow (.json)", then drag that file (or the original PNG) into ComfyUI to load the same graph.

What is PNG Info?

PNG Info is the tab in Stable Diffusion web UIs (AUTOMATIC1111 and Forge) that reads the generation settings saved in a PNG. This page does the same job in your browser, with no install, and also reads JPG and WebP files.

Can I guess the prompt of an image that has no saved data?

Not from the file, because the prompt is simply not there. Some AI models can describe an image and suggest a prompt, but that is a new guess, not the real prompt that was used.

Checking a whole folder?

The full tool lists many images at once with a summary for each.

Open the prompt extractor