AI XMP metadata generator
A dedicated tool to generate Title, Description and Keywords. Supports a local API, experimental WebLLM, manual editing, and bulk updates.
Work fully on-device without worrying about confidentiality
Files are processed on your computer via a local worker, LM Studio, or an API you connect. No cloud upload by default.
WebLLM and metadata settings
Title, Description, and Keywords are generated in the browser via WebLLM. Finished JPEGs embed XMP (no local worker required).
Reference: how the service and export work. WebLLM settings are below.
Cloud / PRO
Generate tags without a powerful GPU. One-click connection, up to ~100 photos/min.
We are collecting a waitlist for early access. Leave your email and we will notify you when cloud mode ships.
File editor
AI XMP metadata generator
Browser-only mode: up to 100 images per session. JPG, JPEG, PNG, WebP. Export: ZIP with JPEG and XMP.
Files for XMP
Selected: 0 / 0
Send files from earlier steps or upload images here.
Duplicate filter
Local near-duplicate search, series grouping, and quick best-frame pick by sharpness.
Quality moderator
Checks anatomy, textures, physics, and artifacts with a vision model before tagging.
XMP generator
Generates Title, Description, and Keywords for agencies with manual edits, bulk edit, and JPEG ZIP export.
How to use the XMP generator
Below is an end-to-end workflow from upload to export. It does not repeat the short help inside the form (browser modes, WebGPU, and file formats).
-
Add images and choose a mode
Drag and drop JPG/PNG/WebP or pick files. For large batches in Chromium, use folder-on-disk mode to avoid the “browser-only” session limit.
-
Configure WebLLM and fields
Pick a supported vision model; optionally open “Extra metadata settings”: length limits, which fields to generate (Title / Description / Keywords), bulk prefixes, and file renaming.
-
Generate and export
Select frames in the strip, click “Generate tags”, wait for completion, edit text manually if needed, then download a ZIP or write to a chosen folder—JPEGs embed XMP (dc:title, dc:description, dc:subject, dc:creator).
Why this tool exists
The generator speeds up microstock metadata prep: the model suggests a title, description, and keyword set from the image, while you keep control of tone and facts.
It is part of the local AI Tools chain: optionally run duplicate filtering and the quality moderator first, then move to XMP so only selected assets get tagged.
By default the site does not upload your files to the cloud—processing runs in the browser (WebLLM / WebGPU) on your machine. The result is JPEGs with XMP for catalogs and editors that read XMP.
What you get at the end
- English metadata drafts aligned with typical stock briefs
- Manual polish and bulk edits before export
- ZIP or folder export with session structure when the File System Access API is available
FAQ
- Does this replace IPTC fields from the camera?
- No: we write XMP inside JPEG (Dublin Core namespaces). Classic binary IPTC IIM is not embedded—for most agencies and software, XMP is enough.
- Can I skip generating only Description, for example?
- Yes: in “Extra metadata settings” uncheck fields the model should not overwrite; they stay as in the card before generation (including empty or hand-entered values).
- Why is the “browser-only” mode limited to 100 files?
- The cap reduces memory pressure for previews, ZIP export, and WebLLM. For larger batches use folder access in a supported browser.
- Do I need the internet after the first WebLLM download?
- The first run downloads weights into the browser cache; further runs on the same model usually work without redownloading until you clear cache or switch models.
- Will metadata work for Adobe Stock / Shutterstock?
- The tool prepares a typical dc:title, dc:description, dc:subject, and dc:creator set; final agency rules and moderation stay with the agency—always review text before upload.