RemoveSubs

Burned-in subtitle removal is a real pain when repurposing clips. The hard part is the background behind the text, especially on moving footage. Does RemoveSubs inpaint each frame independently, or track the caption region across the clip? How does it hold up with high-contrast text over a face or a busy street, and is there a practical clip-length limit before the fill starts to smear?

Burned-in subtitle removal is harder than it looks because the pixels behind the text move every frame. Does RemoveSubs track the caption region across the clip, or inpaint each frame on its own? I would also want to know how it holds up when the text sits over a face.

The translated-captions use case makes a lot of sense: clear the old burned-in line, then add new subtitles in another language. Does it find the subtitle area by itself, or do I mark it? I am thinking of clips where captions jump between the top and bottom of the frame, or switch from one line to two. Also curious what the clip length or resolution limit is on the free tier.