Skip to content

Text on Screen

Find the scenes where a word was shown on screen: a title, a name caption or lower third, a slide, a sign, a scoreboard, a credit. Fast Video Cataloger reads the text in each thumbnail's frame with the text recognition built into Windows, and the Search window finds the scenes by it, the way transcript search finds what was said.

Nothing has to be downloaded. The text is read on your PC; nothing is sent anywhere.

Turn it on

  1. Open Preferences > Video Indexer.
  2. Tick Text on screen: Read the text on screen in every thumbnail.
  3. Pick the Text language. Windows language uses the language of your Windows account. The list shows only the languages Windows can recognise text in; to add one, add the language in Windows Settings > Time & language > Language & region.

From then on, every video you index has its text read. It adds about 30 ms per thumbnail. The text is read from the frame at up to 1280 pixels wide, not from the smaller thumbnail, so name captions and other small text are read too.

The option is off by default. Leave it off if you only want text for some of your videos, and read them afterwards as described next.

Read the text of videos already in the catalog

  1. Select the videos in the Video Catalog or a bin.
  2. Right-click and choose Index > Read Text on Screen....
  3. Choose Selected, Filtered or All videos and click Start.

Each scene's frame is read again from the video file, at full size. When the file cannot be opened (an offline drive, a masked file), the text is read from the thumbnail instead, which only catches large text; the list in the window says so for each video. Text read before for these videos is replaced.

Re-indexing a video with the option on reads its text again too.

To read just some scenes, select their thumbnails in the Video Scenes strip or in Scene Search Results, right-click and choose Read Text on Screen. Only those scenes get their text read again; the rest of the video keeps what it has.

See what was read

Hold the mouse over a thumbnail in the Video Scenes strip or in Scene Search Results. Under the scene's keywords, the tooltip shows the text read in its frame. When there is no Text on screen line, no text was found in that scene, or its text has not been read yet.

In the Video Scenes notes view (Notes above the thumbnails), the text is in a box of its own under the scene's note, to the right of the thumbnail. The box shows only for scenes with text and cannot be edited.

Search for it

  1. Open the Search window, expand Scene Search and tick Scene search.
  2. Type the words in Text on screen contains.
  3. The Scene Search Results window shows the scenes whose frame showed them. Hold the mouse over a scene to see the text that was read in it.

The words are matched whole, like transcript search: artemis finds ARTEMIS, but arte does not. End a word with * to match its beginning: arte* finds ARTEMIS and artwork. Several words find the scenes that showed all of them. Every other filter applies too, so you can, for example, look for a name caption only in the videos of one folder.

What it reads well, and what not

  • Reads well: clean, horizontal printed text such as titles, captions, slides, credits and signs that face the camera.
  • Reads badly: stylised or handwritten lettering, text at an angle or moving fast, text that is small or blurred in the frame, and writing in another script than the chosen language (Cyrillic read with English, for example, comes out as Latin look-alikes).
  • Misread words. Some words come out wrong (Apouo for Apollo) or run together on a busy title. Text that stays on screen is usually read right in another thumbnail of the same shot, so the scene is often still found. If a search finds nothing, try a shorter or different word.
  • One language at a time. Text in another language that uses the same letters is often read anyway; text in another script is not.

On a catalog server

The server reads text on screen while it indexes when EnableTextRecognition is set in its configuration (see ServerConfig.xml), or from Server Setup, in the language set in TextRecognitionLanguage. Windows on the server must have that language's text recognition installed; the server log says when it has not. Sharing a catalog with the Share Catalog wizard copies your settings across.

While you are connected to a catalog server, the Index menu is not there, so Read Text on Screen... for whole videos is not either. Re-index the videos on the server with the option on. Read Text on Screen on selected thumbnails does work: it reads the frame from the video file when this PC can open it, and the thumbnail otherwise.

From the REST API and an AI assistant

The server's scene search takes the words as text: GET /api/v1/thumbnails/search?text=artemis. The AI assistant connector's search_scenes tool has a text parameter, so you can ask an assistant for "the scene with the slide that says Q3 revenue". The scenes they return carry the text that was read, as textOnScreen, so the assistant can quote the slide back to you.