that makes sense, especially the hard FPS floor.
I was thinking of it more like “prefer 12+ FPS”, but you’re right that if the search is allowed to fall through to 8 FPS then the tool can technically succeed while giving the user a worse result than just telling them to trim the clip.
I like the idea of making failure useful instead:
“This won’t fit at acceptable quality. Trim to about X seconds or raise the budget.”
That also solves the duration lesson at the exact point where it matters.
And thanks for the cropdetect tip. I was definitely overthinking auto-crop. I’ll keep manual crop first, but I’ll make the crop data use a simple rectangle so later auto-detect can just prefill the same values.
I’ll run the benchmark once I’ve tested the search changes. If the numbers are noticeably different from yours I’ll post them here.