The problem with a widescreen recording
A screen recording is 16:9, and Instagram Reels, TikTok, YouTube Shorts and Facebook Reels are 9:16. Taking the center strip of a widescreen frame usually misses whatever you are pointing at. A short has to show the thing being explained, large enough to read on a phone.
Step 1: pick self-contained moments
A good short makes sense with no context. Look for a moment with a clear start and finish, one idea, and a payoff within a few seconds: a single tip, one before-and-after, a question answered. Claude reads the transcript and selects four to six such moments from the long recording. A moment that depends on "as I showed earlier" is a poor short, however good it is inside the long video.
Because the shorts come from the same recording as the long video, you do not record twice. What you do change is the opening: a short should lead with the point, not a warm-up.
Step 2: smart panning, following the cursor and the action
The vertical window is about a third of the width of the widescreen frame. Smart panning moves that window across the frame so it stays on what the viewer needs to see. Two signals drive it.
- The cursor log. While you record, a small helper notes where the mouse is and when you click. The camera can follow those positions. It does not record keystrokes.
- The transcript. What you say hints where to look, such as naming a button or a panel.
The pan should move smoothly and not chase every twitch of the mouse. A short that swings around constantly is harder to watch than a plain crop.
Keep the mouse still while you are only talking, leave a beat between sections, and say names of things clearly. These habits make both the pan and the cuts better.
Step 3: check the pan before export
This is Checkpoint 2 of the workflow. Before any short is exported, you watch a preview of each one, with a box on the widescreen frame showing exactly what the vertical clip will display. If the camera should be on a particular area at a particular second, you tell it and the pan is corrected. Fixing a pan at this stage takes seconds; finding out after you posted it does not.
Check three things in every preview. Is the key element inside the box when it matters? Is the text large enough to read on a phone? Does the pan move at a pace that does not make you queasy?
Captions, covers and the legibility problem
Most people watch shorts with the sound off, so captions matter. The workflow produces captions and a post caption for each short, along with a cover image. Text readability is set earlier, at recording time: use Windows scaling and application font sizes so that text is legible when the screen is shrunk to a phone. See the OBS settings page for the scaling advice we use.
What is tested
The panning logic and previews are built and tested on synthetic recordings. They have not been run on a real recording yet, and we will publish what changes when they are.
Frequently asked questions
How many shorts come from one recording?
The workflow aims for four to six moments per long video, but the number depends on how many self-contained moments the recording has. It is a target, not a quota.
What is smart panning?
The shorts are 9:16, but the recording is widescreen. A smart pan moves a vertical window across the wide frame to follow the cursor and the action, instead of cropping the center and missing what matters.
Can I change where the short is looking?
Yes. That is the point of the preview. If the camera should be on a particular part of the screen at a given second, you say so and the pan is corrected.
Do shorts post themselves?
No. The shorts, covers and post captions are saved to a folder, and a person uploads them.