Skip to content
MetaTana

These docs are for MetaTana 2.0, coming soon

On this page

AI Vision

AI Vision helps identify difficult media when filenames, folder context, local NFO, and provider search are not enough.

Pro + credit-metered AI Vision is a Pro feature and uses 1 credit per file each time it returns an answer, even when it can't name the title (a failed request costs nothing). See Use AI Vision for the click-by-click flow with screenshots, and Credits and billing for the cost model.

What AI Vision Receives

The normal AI Vision path uses:

  • sampled frames
  • reviewed filename or title hints
  • media type hints where available
  • year or season/episode context where available
  • compact instructions asking for structured identity output

It never receives the full video file when it identifies a title.

What AI Vision Returns

AI Vision returns identity guidance such as:

  • title
  • year
  • media type
  • optional season and episode
  • how sure it is
  • possible provider ID hints
  • explanation or uncertainty where useful

Provider metadata still comes from metadata providers after identity is known. AI Vision is not the final metadata source for every field.

Credits

AI Vision uses hosted AI credits: 1 credit per file each time it returns an answer, even when it can't name the title (a failed request costs nothing). Asking again about the same file is free.

See Credits And Billing.

Best Use

Use AI Vision for:

  • badly named files
  • ambiguous extras
  • weak episode filenames
  • content with missing local sidecars
  • manual review where visual evidence helps

Do not use it as a substitute for provider IDs when provider identity is already known.

Going deeper Privacy and data for the frame-sampling boundary, and Hosted feature fails when AI Vision refuses with a credit or auth error.