kSocial extension
The image for the website, ready in a single pass.
For every photo on the website someone should rename it, write the alt text in three languages, convert it, resize it for phones and strip the GPS from the metadata. Usually nobody does, and the site is heavy and hard to find. Images for the web does it from the kSocial Library: it looks at the image, proposes name, alt text and metadata, produces the variants and delivers the package.
The client’s photos stay in-house
The analysis is done by Kamzan’s internal vision model, on Kamzan’s servers: the image, scaled down to 1024 pixels, is not sent to any external provider. The code does the rest: true format, weight, dimensions, EXIF metadata. Everything is a proposal you can correct, and you can also just optimise, without generating the SEO. Languages, widths and formats are set once in the Control centre; by default: Italian, English and Spanish, 480, 768, 1280 and 1920 pixels, AVIF, WebP and JPEG. Anyone using it needs the “Images for the web” privilege.
How it works
-
Optimise
Choose the variants’ formats and widths; EXIF metadata is removed if the company wants it. kSocial tells you how many files are created and their total weight, next to the original’s weight.
-
Generate the SEO
The vision model looks at the image and proposes file name, alt text, title and description in each of the company’s languages. Correct what is off and save.
-
Export
“Download zip” with the variants and manifest.json, or “Save to Files” in a folder of your choice, with your Files permissions. If you changed something, the package is rebuilt first.
An image for the website is prepared once, and properly.
What you can do
-
Quality check
PNGs that are really photos, an extension that lies about the format, an original that is too heavy, an image too narrow for the largest width, EXIF with GPS location: each issue with its reason.
-
What the model sees
The subject, the text inside the image and its language, and whether it looks like a stock photo: readers in another language, or anyone seeing yet another handshake, will notice.
-
What is missing
The images that would complement this one, each with a ready-written prompt to copy into a generator.
-
The ready-made snippet
The manifest.json carries the <picture> snippet with AVIF and WebP as sources and JPEG or PNG as fallback, alt text and metadata per language and, if the company wants it, the ImageObject JSON-LD.
-
From wherever you like
From a project’s Library, by dragging a file onto the “Images for the web” page or, with Files active, picking it from a Files folder.
-
Web-ready file names
Lower case, no accents, words separated by hyphens, in the main language: “IMG_2034.png” becomes a name that says what is in the photo, and the variants follow it.
Frequently asked questions
Do the images leave Kamzan?
No. The function accepts only the internal vision model, on Kamzan’s servers: no client photo goes to an external AI provider, even if the company uses one for other functions.
Do I need the AI assistant?
No, it is a separate extension with its own privilege. “Generate the SEO” needs the internal vision model to be active as an engine in Kamzan Account › Artificial intelligence; optimisation and export work without it.
Is the original modified?
No. The original stays in the Library as it is; the variants are created separately. If the original is narrower than the largest width, the widest variant stops at its size, without upscaling.
The other kSocial features
Want to see it on your own plan?
We show you kSocial with your channels and your pillars, not a fake example. Half an hour, and you know whether it is the right place for your editorial plan.
Alex