Skip to content

Translations & AI assistance

Media metadata — labels, alt text, captions — often needs translating, and that's tedious work. Media Directories lets editors translate media inside the browser, and (optionally) hand the first draft to an AI.

The edit modal's Translations panel showing a Spanish name and alt text

Translating media

When a media type is translatable, the edit and upload modals grow a Translations section for its translatable fields (name, alt text, and so on):

  • With one extra language, the translation fields appear inline.
  • With several, each language gets its own collapsible panel, with a "filled" dot so you can see at a glance which languages still need work.

Translations are saved alongside the source item — including for media you're translating as part of an upload.

Availability

Contexts: standalone browser · field widget · CKEditor. Requires core Content Translation and at least one translatable media bundle. Turn the panels on with Enable translations and pick the bundles under Browser settings → Translations.

AI alt text

With the AI submodule configured, image fields get a Generate button next to the alt text. Click it and the configured AI provider looks at the image and writes a first draft you can accept or tweak. It works for media that already exists and for a file you're still uploading, and it's available in bulk across a whole selection.

AI translation

The same submodule adds an AI translate button next to each language in the Translations panel. It fills that language from the source text so a human only has to review, not translate from scratch — again, per item or in bulk.

Availability

Contexts: standalone browser · field widget · CKEditor. Requires the media_directories_ai submodule, the AI module with a configured provider (a vision-capable model for alt text), and — for AI translation — translations enabled as above. Which fields are AI-fillable or AI-translatable is set per bundle on the AI settings tab.

Screenshots need a live provider

Because AI output depends on your provider and prompt, this page doesn't ship canned "after" screenshots — what you get is whatever your configured model returns. See the AI setup page for wiring it up and tuning the prompts.