Reading Cues From Video

Someone hands you a recording of the show with the cue numbers burnt into the corner of the picture. CuePoints can read that text off the video and build your cue list from it, instead of you scrubbing through and typing three hundred cues by hand.

This is the tool for rebuilding a cue list you don't have the file for; an archive recording, a video from another department, a show that was programmed on a desk you can't get an export from.

File > Import CuePoints From File > To Current Track From OCR (Cmd+Shift+O)

If you have the cue list as a spreadsheet or a console file, import that instead; it is exact, and far quicker. See Importing & Exporting. OCR is for when the video is all you have.

How It Works

CuePoints steps through the video, looks at the parts of the picture you point it at, and reads the text it finds there.

You tell it three things:

1. Where to look: draw a box around each piece of text.

2. What each box means: a cue number, a label, a timecode, a fade time.

3. Which cue type the cues should be created as.

Then it processes the video and puts the cues on your current track.

Setting Up The Regions

Click New OCR Regions and CuePoints puts two boxes on the picture to start you off: one for the cue number, one for the label. Drag them onto the real text.

A magnifier follows you as you drag. It appears on its own, you don't switch it on, and shows the picture under the box enlarged, so you can land the edges on the text rather than guessing. Move the whole box and it frames the box; grab a corner handle and it magnifies around the cursor.

Add more boxes for anything else the picture is showing you.

What Each Region Reads

Each region is assigned to one CuePoint Field:

• Cue ID: the cue number.

• Label: the cue name or note.

• Timecode: a timecode burnt into the picture.

• Fade: the fade time.

You only need the ones your video actually shows. A cue display with nothing but numbers needs one region.

Cue ID is the one that drives everything. CuePoints creates a new cue each time that number changes, and whatever the other regions read at that moment goes onto it. So the cue number region is the one worth getting exactly right; a box that clips a digit will read the wrong number all the way through.

If the video shows no fade times, set Default fade time and every cue is created with that.

Choosing The Cue Type

The dropdown at the top left sets which CuePoint Type the new cues are created as, Lighting, Sound, whatever the video is showing you. See CuePoints for setting your types up.

Drawing A Good Box

The reading is only as good as the box.

• Fit the text, plus a little air. Too tight clips characters; too loose picks up whatever is next to it.

• One box per piece of text. A single box over a cue number and its label will read them as one run of text.

• Use a frame where the text is clearly visible to place the boxes. Scrub to a moment mid-show rather than working on a title card or a fade.

• Check the text doesn't move. If the display shifts position partway through the recording, process the video in sections and reposition the boxes for each.

Processing

One setting: how much of the video to run.

• First 30s: a quick test.

• First 5 min: a longer check.

• Process All: the whole thing.

Run First 30s before you commit to the whole video, every time. In a few seconds you learn whether the boxes are on the right text, rather than finding out at the end of a two-hour pass. Check the cues it produced, adjust the boxes if you need to, and only then run Process All.

Press Process Video to start. Abort stops a run you don't want to wait out.

Reading a full show takes a while, so it is worth starting Process All when you can leave it to get on with it. When it finishes, the cues are added to your current track.

Templates

The same rig films the same display in the same corner every time. A template saves your regions, their fields and the cue type, so next time is one click.

• Save Template… stores what you have set up, under a name.

• Pick a saved template from the dropdown to load it.

• Revert goes back to the template as it was saved, throwing away your changes.

• Reset Boxes puts the regions back to the default pair.

• Delete Current Template… removes the selected one. This cannot be undone.

If you regularly get video from the same source, build a template once and this becomes a two-minute job.

Checking The Results

Look at the cue list before you do anything else. OCR reads what it sees, and a burnt-in display can be blurry, compressed, or briefly covered by something in front of it.

The things to check:

• The first and last cues. If those are right, the boxes were in the right place for the whole run.

• Numbers that jump.101, 102, 108, 104 means one frame read badly. Fix it by hand.

• Doubled cues. A number caught mid-change can read as two. Delete the stray one.

• Anything at the very start. Title cards and fades produce nonsense more often than the body of a show.

A pass that gets ninety-five per cent of a three-hundred-cue show and needs a few fixed by hand has still saved you a day.

When It Isn't Working

Nothing is found. The boxes aren't on the text. Scrub to a frame where the display is clearly visible and check the boxes sit over it there; the frame you set them up on may not have been showing it.

The text reads as gibberish. The box is probably too tight and clipping the characters, or it's picking up something alongside the text. Widen it slightly, and make sure it covers one piece of text only.

Cues are missing. The number wasn't readable in those moments. Scrub to one of the gaps and look at the picture; the display may be blurred, briefly covered, or fading. Widen the box slightly if it is clipping, and add the missing cues by hand.

Cues stop partway through. The display moved. Process the video in sections, repositioning the boxes for each one.