Drop your images here
or click to choose multiple files
PNG / JPEG · all parsed locally · processed in your browser · never uploaded
Drop or click to replace
Add PNG/JPEG images to extract their prompts and settings into one table.
A tool that reads the Stable Diffusion generation data embedded across a whole folder of images and lays it out one row per image. It extracts the prompt, negative prompt, seed, sampler, steps, CFG scale, model name (hash), and any LoRAs used from each image, then lets you export the lot as CSV — ready to open in Excel, Google Sheets, or pandas. Instead of inspecting images one at a time with PNG Info, this is for when you want the whole picture: list every seed and setting in an output folder, compare prompt patterns across a set of favorites you collected, or review which training images or wildcards worked at a glance in a table. The columns are gathered automatically from whatever the loaded images contain, so all the various A1111 parameters become columns as-is. Because prompts are often as private as a recipe, this tool never uploads your images — both the parsing and the CSV generation happen entirely in your browser.
How to use
- Drop your images in together, or click to choose multiple files (PNG / JPEG / WebP).
- Each image's prompt, negative, settings (seed / sampler / steps / CFG / model), and LoRAs are lined up in one table.
- Use "Download CSV" to save the table, or "Copy CSV" to paste it straight into a spreadsheet.
FAQ
Are my images uploaded to a server?
No. Both the metadata parsing and the CSV generation happen in your browser. Your images are never uploaded, stored, or sent anywhere — everything is processed on your device, so it's safe even when handling a large batch.
How many images can I process at once?
There is no hard limit. Each image is read only in its metadata region (pixels are never decoded), so hundreds of files stay snappy. Loading is done incrementally so the UI doesn't freeze.
Which columns appear in the CSV?
Filename, source, prompt, negative, and LoRA, plus every setting found in the images (Steps / Sampler / Seed / CFG scale / Size / Model and so on) collected into columns automatically. Even if images differ in which fields they have, the columns are aligned across the whole set.
The Japanese (or accented) text is garbled in Excel.
The downloaded CSV is written as UTF-8 with a BOM to avoid this. Excel should open it directly; if it still looks wrong, use Data → From Text and choose UTF-8 when importing.
Which generation tools does it support?
AUTOMATIC1111 (A1111) WebUI PNG info, ComfyUI embedded workflows, NovelAI metadata, and JPEG / WebP EXIF UserComment. To inspect a single image in detail, use the PNG Info tool.
Why are some rows empty?
Those images probably have no generation info embedded. Metadata is commonly stripped by social media uploads, re-saving in an editor, screenshots, or compression/conversion — once removed it can't be recovered. Images with no info are still kept as rows so you can see which ones are missing.