How Split framing works
When you choose Split for a clip, MyShorts samples frames from that stretch of the episode and finds faces. If the faces fall into two groups clearly apart from left to right, it treats them as two people. It then cuts two panels, each 1080 pixels wide and 960 tall, centres each one on its person's face, and stacks them into a 1080 by 1920 frame.
Each panel is only as wide as the space between the two people, so a panel never spills over and shows the other person's shoulder. The positions are set once for the clip, which keeps the two panels steady instead of drifting with every head movement.
| Part of the frame | Shows | Size |
|---|---|---|
| Top half | The person sitting on the left of the wide shot | 1080 × 960 |
| Bottom half | The person sitting on the right | 1080 × 960 |
| Captions | Word-timed text, low in the frame by default; move it to sit on the seam if you prefer | Adjustable |
When Split falls back to Speaker
Split won't fake a second person. It switches to Speaker framing (following one face) when:
- only one person is on screen for most of the clip, as in a close-up from a multi-camera edit;
- the two people sit so close together that their faces don't form two separate groups;
- faces are too small, turned away or hidden behind mics for the detector to find enough of them;
- there are no faces at all, as with cover art or slides.
If your clip came out as a single-face crop when you asked for Split, one of these is usually why. Try a moment where the wide shot holds longer.
How to film so Split works every time
- Use one locked-off wide camera that sees both people from the chest up.
- Leave a clear gap between the chairs; roughly a chair's width is enough.
- Keep mic arms low or to the side so they don't cover mouths and chins.
- Light both faces evenly; a dark side of the table makes one face harder to find.
- If you also shoot close-ups, keep the raw wide-camera file. Put it on YouTube as unlisted and clip from that for the cleanest Split results.
Split, Speaker or Wide: choosing per clip
| Moment | Framing | Reason |
|---|---|---|
| Quick back-and-forth, both people reacting | Split | Viewers see the question land and the answer start |
| Guest tells a two-minute story | Speaker | One face, larger on screen, more emotion visible |
| Host and two guests laughing together | Wide | Split handles two people, not three |
| Interview over a video call shown side by side | Split | Works if the two windows are far enough apart |
You set framing on each clip, so one episode can give you a Split clip of an argument and a Speaker clip of a story.
Captions on a stacked frame
With two faces taking the full height, the caption position matters. The default sits low, over the lower speaker's chest. Some shows move captions to the middle, across the seam, so neither face is covered. Podcast Box with its dark card reads well on either panel; Pill Highlight works if you keep it to one line. On Free the choice is Basic and Basic Box; paid plans from ₹299/month unlock every template.
A good Split clip, described
- Hook (0–3 s): host asks, “Would you do it again?”
- Guest pauses, looks down. Top panel shows the host waiting.
- Answer (3–40 s): “No. And here's the part nobody tells you…”
- Captions: Podcast Box, two lines, centred on the seam.
Read the full guide: Face-aware 9:16 reframing: keep the speaker in frame →
Last updated: