Make a Lens that edits any photo

AI Edits takes one written edit and builds a Lens that applies it to whatever photo someone takes. You are not editing your own picture, you are writing the instruction that gets applied to theirs.

Open the Gen AI suite at /assets. AI Edits leads the grid, on the card that reads Describe the edit you want — a new style, a new place, a new look — and AI Edits builds a Lens that applies it to any photo, and Try Now opens it. There is a second way in at /ai-edits, which puts the same box on a page of its own.
Read the header before you type, because it is the part people get backwards. Edit Any Photo With One Prompt is what the tool does, and the line under it says who the photo belongs to: Describe one edit — your Lens applies it to any photo a Snapchatter takes. The line beside the icon is what it costs you, an Average Processing Time: 15 seconds.
Pick a Category before you write anything. All is what you land on, and the seven beside it — Selfie, Scenes, Style, Lighting, Outfits, Camera, Characters — each swap the five previews under the box for that kind of edit. The five you can see are the start of a much longer list, so Shuffle Concept keeps dealing new ones from whichever category you are on rather than cycling those five.
Use /ai-edits if you are not sure which category your idea is. The same seven run along that page under One prompt box, every kind of edit, and each one comes with a line saying what it is for — Selfie is Keep the photo, change the look — a new cut, a new colour, or full-on makeup — with worked examples under it and a Try this on each that carries the prompt straight into the box.
Write one edit in Concept, which is the only thing the tool requires. It asks you to Describe the edit — a new style, outfit, lighting, or scene, and one edit is the unit: "Make it look like a 1970s film photo, warm grain and faded colour." A concept that asks for four unrelated changes at once gives the model nothing to hold constant, and holding the rest of the photo constant is the whole trick.
Read the line along the bottom before you build. It prints the sentence that actually gets sent, Create a Lens using AI Edits of… with your concept on the end. If that sentence does not describe the edit in your head, fix the concept rather than the Lens.
Tap Create My Lens, then change one thing per message. The page's own worked example is three messages rather than one: "make ai edits lens to make me an ogre", then "give the ogre a cute pink sparkly armor with little toys & details", then "ogre face is too serious now, make it more cute and add blush". Each message keeps what you did not mention.
Tap Publish when the edit is right, and review the name, icon and description Easy Lens writes before you submit. Once it is live, anyone can try that edit on their own photo, which is the thing you were actually building.

You have a Lens that applies one described edit to whoever opens it, and you know to fix the concept rather than the Lens when the sentence at the bottom is wrong.