Images
Get the picture, and everything after the picture.
Make one, bring one back from the web, or upload your own — then the description, the crop and every size a browser might want are handled without you. The tedious part is the part that gets skipped, so it is the part that is automatic.
Three ways in
Generate one from a description, import one off a page the AI read, or upload your own.
Described automatically
Alt text written by a vision model and a focal point detected on upload. Both free, both editable.
Served properly
The right size and the smallest format each visitor can read — never the full-size original.
Getting a picture
Three routes, one library.
However a picture arrives it lands in the same place, gets the same treatment and is available to every page and to the AI. There is no separate “generated images” bucket to manage.
Describe it
8 shapes and 3 sizes, each saying plainly what it is for — so a small logo is not made at the largest size and charged like a full-width photograph. It can also work from up to 3 pictures you already have, so a new one matches the ones around it.
From a page it read
Point it at a page and it can pull a picture off it into your own library — a real copy that belongs to you. Simply linking to somebody else’s image gets you none of the handling below, and it vanishes the day they reorganise their site.
Yours
Drag the folder in. The progress bar shows what has genuinely been sent rather than a guess — a big photograph on a slow connection is exactly the moment a made-up bar gets caught out.
What happens next
The four jobs nobody does by hand.
A description, written for you
Every picture needs a short description of what it shows — it is what a blind visitor hears read aloud, and what a search engine reads instead of looking. The AI looks at the picture and writes one, free, the moment you upload it, and you can change it. It is the most commonly skipped field on the web, and never because people do not care: it is asked for at the exact moment somebody is trying to finish something else.
Where the subject is
The same photograph has to be a wide strip on a laptop and a tall crop on a phone. Knowing where the subject actually is means the crop keeps it. Cutting to the middle of the frame — which is what happens by default nearly everywhere — is why so many websites take the tops off people’s heads on a phone.
The smallest version that works
Newer picture formats are dramatically smaller for the same quality — measured at 34–44% smaller, and indistinguishable side by side even zoomed in. Each visitor’s browser is offered the choices and quietly takes the best one it understands, so nobody on an older phone is left without a picture.
Sizes that really exist
A phone is sent a phone-sized picture and a large monitor a large one, made when first needed and then kept. Crucially, a size bigger than the original is never offered — telling a browser a small photograph is available huge is the common mistake that downloads the most and displays the worst.
The library
One screen, and it turns up wherever you need it.
It is the same library everywhere you need one — choosing the picture at the top of a page, setting the little icon in the browser tab, or attaching a screenshot to a message for the AI. Upload and make-a-new-one are inside it, so you never have to leave and come back.
Search by what is in it
Search by filename, by the description written for you, or by what was recognised in the picture. “team”, “storefront”, “shopfront at night” — you find the photograph without remembering what the camera called it.
See what uses what
Every file knows which pages, blocks and settings reference it. That is what makes deleting one safe to offer: you are shown the damage first, by name, and then allowed to do it anyway.
Logos stay crisp
A logo saved as a vector file is served as one, at any size, with nothing resampled. It is also cleaned on the way in — that file type is the one that can carry code, and the people most at risk would be your own visitors.
Questions about images
What does a generated image cost?
5 to 12 credits depending on the size you pick. The price is shown before you make it, so nothing is a surprise. Uploading your own and bringing one back from the web are both free.
Do I have to use generated images?
No. Upload your own, or have the AI bring one back from a page it read. An imported picture is treated exactly like one you uploaded — same sizes, same description, same everything — except that it remembers where it came from, which is the question that comes up about it later.
Who writes the picture descriptions?
The AI looks at the picture and writes it when you upload, free, and you can change it afterwards. It is the short description a blind visitor hears and a search engine reads, and it is the field most commonly left blank on real websites.
What is a focal point?
Where the subject is. It is detected automatically when a picture is uploaded, and it decides what survives when the same photograph has to be a wide banner on a desktop and a tall crop on a phone. Without it, cropping to the centre cuts people’s heads off — which is the default behaviour almost everywhere.
Can it edit a photo I already have?
It can make a new picture working from up to 3 you already have — “another like this one”, “the same style as the one at the top”, “this product on a plain background”. It is making something new rather than retouching: you get a new picture and the original is untouched.
What happens if I delete a picture that is being used?
You are told which pages, blocks and settings use it, by name, before you confirm — and then it is deleted, because it is your file. Most tools either silently break the pages or refuse outright; being told and then allowed to proceed is the useful middle.