Prompt Metadata Checker for Stable Diffusion Images

Reads the metadata of an AI generated image (using the Exif or tEXT enclosed in that image file). Populates the prompt, seed, sampler, and more. Known to work well with images generated using AUTOMATIC1111, InvokeAI or NMKD. Supports both .png and .jpeg images.

If you run into any issues: please open up an Issue here on Github or talk to us on Discord.

This is a port into PromptHero of a community extension originally made by @HE1CO and later revamped by @drhino

Motivation

This is what we use in our prompt upload page. When you drop a valid image generated with A1111 or a supported program, this is what auto-fills all the fields in the page.

Contribute

If you would like for PromptHero to support reading the metadata of other file formats or images generated by other programs, please feel free to open a pull request!

How to develop

The heavy-lifting work is done in the JS files found under src/vendor/postie. That's where most of the original browser extension's code has been imported.

The src/controllers folder holds the Stimulus controller that manages the user interface found in index.html

JavaScript code is imported/managed using Import Maps. The project importmap is found in index.html

To develop, just open index.html in a web browser and edit the relevant JS files. No JS tooling required.

Features

• Automagically fills upload info in the form after you choose a file
• Works with images generated by Stable Diffusion web UI, InvokeAI, and NMKD

Limitations

• NMKD: The model hash is not saved in the PNG metadata, so the model name and version can't be auto-filled
• General: Inpainting overwrites the original prompt info, so the extension will autofill the inpaint prompt