Skip to content

AI tool

AI Metadata Generator

Gemini reads the image itself and writes an SEO title, a description and keywords ordered by relevance, checked against Adobe Stock and Shutterstock limits.

Upload your image

Drag and drop, browse, or paste with Ctrl/Cmd + V. EPS is read via its embedded preview.

How it should be written

Set these before generating — they shape the title length, the keyword count and the wording.

0 / 400

Steer the wording, e.g. “emphasise autumn, aimed at travel blogs”.

About the metadata generator

Upload an image and Gemini describes what is actually in the frame, then writes a stock title, a description and a keyword set from that description. It reads the picture rather than the filename, so an image straight off a camera gets the same treatment as a carefully named one.

The output is a draft — a good one, but a draft. You know things the model cannot see: where it was taken, what the shoot was for, which concept the buyer is likely searching for. The review pass is where a submission becomes yours.

When this is the right tool

  • Preparing a single image for a stock agency and starting from a blank field.
  • Working through a backlog where writing metadata is the bottleneck rather than the shooting.
  • Looking for the keywords you would not have thought of — the concept and mood layer people usually miss.
  • For a whole folder rather than one image, use the bulk generator instead.

How to use it

  1. Add your Gemini API key once

    Free from Google AI Studio, stored in this browser only. Paste it into Settings and you will not be asked again on this device.

  2. Upload the image

    It is downscaled in your browser before being sent, because vision models tile their input anyway — a full-resolution upload costs time without improving the result.

  3. Set your options

    Title length, keyword count within the 30–49 window, words to exclude, and an optional prefix or suffix. These are remembered on this device.

  4. Review, then export

    Read the title aloud and delete any keyword you cannot point at in the frame. Then export as TXT, or as a CSV shaped for the platform you are uploading to.

Requirements and limits

Input
JPG, PNG and WebP up to 10MB, plus EPS up to 45MB — an EPS is read through its embedded JPEG preview, since no browser can decode PostScript directly.
Keywords
30–49, ordered by relevance, within a title capped at 200 characters.
What is sent
A downscaled copy of the image, forwarded through this site's own route to Google rather than from your browser, so the key never appears in a page URL. It is not logged or stored; the request is held in memory and discarded.
What it cannot do
It has no access to any agency's search or sales data, so it cannot tell you what will sell or how a submission will rank — what to judge instead.

Common questions

Do I need to pay for anything?

No. The tool is free and the Gemini API key it uses is free too — Google AI Studio issues one without a billing account. The free tier is rate limited rather than charged, which is why the bulk tool paces itself.

Where is my API key stored?

In your own browser's localStorage, on your device. It is sent with a request only when you generate something, and only to Google via this site's own route handler. It is never logged, never put in a URL and never shared with anyone else.

Will the title and keywords fit Adobe Stock and Shutterstock?

Yes. Titles are capped at 200 characters and keywords at 49, which are Adobe's limits — a list that fits Adobe also fits Shutterstock's 50, so one set of metadata is submittable to both. Both limits are sliders if you want something shorter.

Does it look at the image or just the filename?

The image itself. The file is sent to Gemini as an image and the model describes what is actually in the frame, so a photo named IMG_4821.jpg gets the same quality of metadata as a carefully named one.

What can I export?

A plain TXT file, or a CSV shaped for the platform you are uploading to — the column names and ordering differ between agencies, so the export matches the one you pick rather than making you rearrange it by hand.

Is there a file size limit?

10MB per image. That is well above what a JPEG preview needs — you do not have to upload the full-resolution master to get metadata for it.

Guides on this

Longer answers to the questions this tool raises.

Related tools