Everything it does, in one list.
Free unless marked Pro. Pro is $15 once, on three browsers, with lifetime updates.
Getting an image in
Anything you can see is fair game.
Right-click any image
The context menu on any page, including images inside iframes.
Hover badge
Appears on images over 128px. Lives in a closed shadow root, so no site's CSS can break it and it cannot break any site.
Screen region
Drag a box over canvases, video frames, maps, PDFs and CSS backgrounds. Needs no permission on the site at all.
Drop, paste, or pick
Files from disk or the clipboard, straight into the panel.
Whole page at once
Finds every image over 96px including CSS backgrounds, biggest first; you tick the ones to read.
Side panel or tab
Works beside the page you are reading. Falls back to a tab on older Chrome rather than failing.
The reading
Ten dimensions, not a caption.
Structured analysis
Subject, scene, composition, camera, lighting, colour, texture, mood, style and any lettering — plus a negative list of what must not appear.
Measured palette
Six dominant colours read off the pixels in your browser, named and hexed, with warm/cool, saturation and key derived from them. Free, offline, and more accurate than asking a model.
Measured aspect ratio
Taken from the real dimensions and snapped to a ratio a generator accepts, so --ar is never a guess.
Style-only mode
Reads the look and deliberately leaves the subject blank — for borrowing a treatment, and for images with people in them.
Subject-only mode
The reverse: what is in the frame, with the styling stripped out.
Fifteen languages
Prompt values in English, Welsh, French, German, Spanish, Italian, Portuguese, Dutch, Polish, Turkish, Japanese, Korean, Chinese, Arabic or Hindi.
Getting a prompt out
Nine formats and a dial, all from one call.
Plain English
Free. Readable by a person, pasteable anywhere.
Eight more dialects
Midjourney flags, SDXL weights and negatives, a Flux sentence, a DALL·E paragraph, an Ideogram lettering-first prompt, a Gemini instruction, a video shot with a camera move, and raw JSON.
Faithfulness dial
Loose inspiration, balanced, or verbatim recreation. Constraints like the negative list and the aspect ratio are never trimmed away.
Eight variations
Night, golden hour, studio, pull wide, push macro, black and white, as illustration, on 35mm film. Rewritten locally — no second API call.
Compare and blend
Take the subject from one image and the lighting, grade, texture and lens of another. Plus a field-by-field diff of the two readings.
Send-to links
Open Midjourney, ChatGPT, Gemini, Ideogram or Leonardo with the prompt already on your clipboard.
Keeping it
A library, not a scroll-back.
Searchable history
Every field and every generated prompt is searchable, so a colour finds a shot whether it came from the palette or the grade.
Favourites survive trimming
Starred readings are kept when older entries fall off the cap.
Export
JSON, CSV or Markdown. Thumbnails are stripped from every export, because a prompt library is text.
Local only
Nothing syncs, nothing uploads, and Clear all is the whole delete.
Cost estimate up front
The panel shows the rough per-image cost for your provider before you press the button — and says "free" for local models.
Test-this-setup button
Proves the key, endpoint, model and reply parsing in one 2×2-pixel call, so failures surface on the settings page rather than mid-job.