What cropping actually costs
A 16:9 clip with a caption in the lower tenth needs that tenth removed, then the remaining nine tenths stretched back to 1920×1080. Every pixel is resampled, so detail across the entire frame loses sharpness — not only where the text was.
Cropping also changes composition. If the caption sat over a subject’s hands, or over a lower-third graphic, that content is gone for good. And unless you crop to a different delivery ratio, the aspect ratio no longer matches the rest of your edit.
- Reframing is permanent — the cut pixels are not recoverable from the output
- Rescaling softens the whole frame, not just the repaired strip
- Aspect ratio changes unless you deliberately crop to another format
Where cropping is still the right call
If the caption sits in a strip that contains nothing you need — a letterboxed bar, a solid colour band, the bottom of a wall — cropping three percent and keeping the rest is cheaper and faster than any reconstruction, and it cannot introduce artefacts.
The rule we would use: crop when the strip is genuinely empty, reconstruct when the strip contains picture worth keeping. Most subtitle removers only offer the first option, which is why this page exists.
How far the reconstruction reaches
The model only touches the area you mark. A tight box around the caption band leaves the rest of the frame as it was, apart from the single re-encode the export requires.
On textured backgrounds the repaired strip is softer than its surroundings. At normal viewing size that reads as motion blur; at 100% you can find the patch. That is the trade for keeping your framing, and it is a smaller trade than losing the edge of every frame.