Concretely, the frames are scaled with a high-quality filter and the stream is re-encoded, since resolution is a property of the coded picture; the aspect ratio is preserved unless you deliberately override it. You give it a target width, a target height or a percentage, and it works from the original rather than from a previously resized copy.
Can I keep the aspect ratio, or force exact dimensions?
+
Both. Lock the ratio and one dimension follows from the other, which is what you want almost always. Unlock it and you get exactly the numbers you typed, at the cost of stretching — which on video means every face in the frame is slightly too wide. When you need a different shape, crop or letterbox rather than stretch.
Does the source format affect the result of a resize?
+
It does. MPEG (MPEG-1/2) is the DVD and broadcast lineage: robust and universally decodable, but far less efficient per bit than H.264. That is what decides whether shrinking the file is purely a question of pixels or also a question of re-encoding.
Is there anything specific to MPEG I should watch for?
+
Yes — MPEG-1/2 uses a program stream rather than a modern index, so cuts are aligned to GOP boundaries unless the picture is re-encoded. Worth checking before you resize a file you will not be able to regenerate.
What is the most common mistake when resizing MPEG videos?
+
Upscaling. Enlarging cannot invent detail that was never captured, so the result is a soft version of the original at a bigger size. Always resize down from the largest source you have, and resize once rather than in stages.
What resolution should I actually target?
+
Match where it will be played. 1920x1080 is still the safe default for almost everything; 1280x720 roughly halves the bitrate and is plenty for a screen recording or a talking head; 3840x2160 is only worth keeping if the footage was genuinely shot at it. Vertical platforms want 1080x1920. Never scale up.
