Leave Room for the Reel UI
Keep the bottom and the right clear. Captions and the face sit in the upper-middle, where the reel buttons do not cover them.
Leave the bottom and the right clear for the reel interface. Jun's face sits in the upper half, and the caption sits upper-middle, which is the safe zone. Likes, comments, and the app's own caption bar occupy the bottom and the right. If the mouth or the words sit there, the interface covers the point. One person speaks. Burned-in captions help when the same file is watched on TikTok or Shorts with the sound off. A realistic face or voice is labeled as AI. This framing is decided in the still, before the mouth moves. Export only after the crop is safe: Jun's eyes in the upper half, the bottom and the right left empty, the closing line in the upper-middle. The first three seconds show his face high and the words clear of the buttons while he says the bakery closes at four. A mouth on the bottom bar, or type on the right where the buttons sit, is the frame you reject before posting. One speaker, fifteen to forty-five seconds, and the caption says the face and voice are AI.
Park the face high enough
Jun stands so his eyes fall in the upper half and his hands do not stretch into the right edge. Hearth Bakery's counter can run along the lower part of the frame as empty wood. A face cut off by the top, or a mouth sitting on the bottom bar, fails the still. Regenerate the crop before you approve speech. Stand him so the eyes land in the upper half, keep the hands off the right edge, and regenerate the crop before any speech. The first three seconds show the face high, with empty wood along the bottom, while he says they close at four. A face sliced by the top edge, or a mouth on the bottom bar, is discarded.
Put words upper-middle
The caption safe zone is the upper-middle. The line about closing at four belongs there, in a caption style you can read at phone size. Words along the bottom collide with the app caption. Words on the right collide with the buttons. Burned-in captions stay in that upper-middle band so the file still reads on TikTok and Shorts. After the face is framed, burn the hour in the upper-middle, in a caption look you can read on a phone, so a silent play on TikTok or Shorts still shows it. The opening three seconds show those words clear of the buttons while Jun says they close at four. Words along the bottom, or stacked on the right, are the layout you reject.
One speaker, labeled
Only Jun is in the shot, so the empty zones stay empty. A second person walking into the right edge is a framing error. Because the face and voice are realistic, label the post as AI. Leave the platform label where the app places it. The practical cut of fifteen to forty-five seconds uses this same frame. One person keeps the empty zones empty. Label the realistic face after the crop passes, and export this same frame for the whole practical cut. The first three seconds show that single speaker high in the frame while he says the bakery closes at four. A second person walking into the right edge is refused before posting.
A worked example
Jun at Hearth Bakery is framed with his face high, counter wood along the bottom, and the right side quiet. The line about closing at four is captioned upper-middle. The post is labeled as AI. A still that put his hands on the right edge and the words along the bottom was rejected before he spoke. Face high, bottom clear, line upper-middle, then speech. That is the order. The first three seconds show Jun with counter wood along the bottom while he says they close at four. The still that put his hands on the right and the words along the bottom is the one you throw out. The post says the host is AI.
Before you post
- The bottom and the right are clear.
- Captions sit upper-middle.
- The realistic host is labeled as AI.
Where Scro fits
Scro keeps one person speaking and offers 13 caption styles for the line in the upper-middle. Export at 9:16. Label a realistic face or voice as AI, and leave the platform disclosure on the post. Approve the safe-zone still, set one of 13 caption styles in the upper-middle, and export at 9:16 after that. The first three seconds show the line clear of the reel buttons while Jun says they close at four. A covered mouth is rejected before posting. The caption says the face and voice are AI.
Clear the bottom and the right. Finished samples are on the examples page.
Related: Approve Stills before Motion.