Describe new images on upload with AISA Vision

What it is

Describe new images on upload is the automatic side of AISA Vision: instead of opening every image you have just uploaded and asking for a description by hand, AISA describes it on its own a few seconds after the upload. It covers every way an image enters WordPress — the file picker, drag and drop, the Media → Add New screen, the block editor — because they all pass through the same WordPress event.

The option is off by default, on purpose: every image that gets described uses tokens with your API key, or about one credit on AISA Cloud. While the switch is off nothing is described, no background job is scheduled and nothing is spent.

It is the same engine as “Describe with AISA Vision” in the Media Library: same prompt, same cost, the language taken from the real context of the image, the “do not translate” glossary respected, and the image title written the WPML-safe way.

Where to find it

AISA → Settings → API Keys, in the AISA Vision section, right under the image provider. The section holds the switch, the choice of what to write, the safety option, a cost line calculated on the provider you actually have configured, the last-24-hours status line and the Describe queued images now button.

If no image provider with Vision is configured the checkbox stays disabled, with a link that reads Set an image provider with Vision first: a toggle that looks on and does nothing would be worse than a disabled one.

Requirements & tier

Free, on every tier: it runs on your own key like the rest of Vision, or on credits with AISA Cloud (about one credit per image). It needs an image provider with Vision configured (Claude, GPT or Gemini), the upload_files capability for whoever is uploading, and WP-Cron for the background job — if WP-Cron is disabled the images simply wait in the queue and the button empties it.

Three things are always skipped: files that are not images (a PDF is never described), SVG files (a vision model on a vector graphic gives unreliable results), and the images AISA generated itself, such as a social/OG image or a video poster.

How to use

Open Settings → API Keys and check that an image provider with Vision is configured.
In the AISA Vision section tick Describe new images on upload.
Choose what to write: Alt text only (the default) or Alt text and title.
Leave Only when the alt text is empty ticked if you sometimes type the alt yourself in the upload dialog: with it on, what you wrote is never overwritten.
Save, then upload an image as usual. The upload dialog shows AISA Vision will describe this image in a few seconds, and the alt field fills in on its own when the description lands.
Come back to the section whenever you want to read the status line: Last 24h: N images described, N skipped (alt already set), N errors, plus queued: N when something is still waiting.

For safety no more than 50 images an hour are described automatically. Past that the images stay in the queue — nothing is lost — but a catalogue import of hundreds of photos belongs in Bulk SEO → Images, which shows the cost estimate before it starts.

Questions people ask

Does it cost anything if I leave it off? No. With the switch off no image is described, no background job is scheduled and no call is made: the option costs exactly nothing until you turn it on.
I uploaded an image and the alt field is still empty. Give it a few seconds: the description runs in the background right after the upload, and the dialog fills the field on its own. If it stays empty, WP-Cron is probably disabled or slow on your site: the status line will read queued: N and the Describe queued images now button processes them five at a time.
I upload 300 product photos at once. What happens? The first 50 within the hour are described automatically; the rest wait in the queue, and the section says so. For an import that size use Bulk SEO → Images: it runs the whole catalogue and shows the cost estimate first.
It wrote the alt in the wrong language. The language comes from the real context of the image — caption, title, the page it belongs to, then the site language. Fill in the caption or the title before uploading, or describe the image again from the Media Library, where you also have Undo.
Does it overwrite the alt I typed in the upload dialog? Not with Only when the alt text is empty ticked, which is the default: an alt you wrote yourself is left alone and counted as “skipped” in the status line.
I also use an image optimizer (WP Compress, ShortPixel…). They work together. Optimizers compress and rewrite the image file after the upload; AISA Vision only writes the alt text, and it waits for the optimizer to finish before looking at the picture, so it never describes a file that is still changing. The one real clash is with a plugin that writes the alt itself — Rank Math’s “Add missing ALT attributes”, All in One SEO’s Image SEO and similar: whoever writes first wins, and since they fill the alt from the file name, AISA finds the field already taken and leaves it alone. Turn that option off in the other plugin and AISA Vision writes the real description. The AISA Vision section lists what it detects on your site, so you know which case you are in.
The provider failed: where do I see it? Never on screen while you are uploading. Errors go to AISA → Utility → Error log with their cause, and they are counted in the errors figure of the status line.

AI alt-text with Vision
Describe with AISA in the block editor (Gutenberg)
AI image titles