Remove subtitles from video clips you own or are allowed to edit, clearing burned-in captions and other on-screen text that are part of the picture itself. Upload one source clip, check the credit estimate, and let the AI rebuild the area behind the lettering across the sequence.
Hardcoded subtitles are pixels in the frame, not a separate track, so an ordinary editor cannot switch them off. The workflow detects the lettering on its own, which spares you from drawing masks or tracking caption boxes line by line as they change through the clip.
Try it nowInstead of blurring or cropping away the lower part of the frame, the model reconstructs the background behind each line using nearby detail and neighboring frames. Plain backgrounds usually recover cleanly, while busy textures and fast motion deserve a closer look.
Try it nowTiming, cuts, camera movement, and the on-screen performance stay as they were in the source. Only the text regions are processed, so the cleaned clip drops back into your timeline without re-syncing audio or re-cutting scenes.
Try it nowOnce the old lines are gone you can add translated subtitles, corrected wording, or a caption style that suits another platform, without the new text fighting the old text underneath it.
Try it nowDialogue lines rendered into the frame during export, including translated subtitles baked into an older version of your video.
Auto-generated or styled captions that a short-form editing app added before the clip was exported and saved.
Name titles, location labels, and other text strips that sit over the footage rather than on a separate editing layer.
Recording dates and time codes printed into camera or screen-capture footage that you now want to reuse elsewhere.
Lines that stay on screen while the camera pans or the subject moves, which are tedious to patch manually one frame at a time.
Clearing the original language from the frame so a new audio track and matching subtitles can be added in your editor.
Reuse your own older posts in a new crop or format, where the original captions no longer line up with the new framing, pacing, or hook.
Refresh licensed or in-house campaign footage whose captions carry an outdated offer, price, or date before it runs again.
Clean your course recordings and walkthroughs so the same lesson can be relaunched with subtitles in another language.
Prepare client-owned footage for dubbing or re-subtitling when a clean, text-free master is no longer available.
Add one video you own or have permission to modify. The least-compressed version you have gives the model more real detail to work from.
Review the credit estimate beside the Generate button, then submit. Text detection runs automatically, so there are no caption boxes to mark by hand.
Play the whole result, watch the areas where the subtitles used to sit, and download the clean version once you are satisfied with it.
Only upload videos you own or have permission to modify.
Trim the clip to the section you need before uploading; billing follows the length of the source.
Use the original export instead of a re-downloaded copy, since heavy compression leaves less detail behind the text.
Large, thick, or boxed captions cover more of the frame and are harder to rebuild than thin single lines.
Check faces, hands, and patterned backgrounds that sat under the subtitles, and keep your original file for comparison.
It means erasing text that has been burned into the picture, such as hardcoded subtitles, captions, and similar overlays, and filling that area with reconstructed background. Soft subtitles stored as a separate track can simply be turned off in a player or editor and do not need this tool.
Only when you own the footage or have permission to modify it. Clearing the captions does not give you any rights to someone else’s video, and redistributing it may still infringe the creator’s copyright.
It is built for on-screen text: subtitles, captions, titles, labels, timestamps, and similar overlays. It is not a general object remover, so people, props, and other visual elements in the scene are left in place.
It depends on what lies behind the text. Plain skies, walls, and slow shots usually reconstruct well. Large caption boxes, fast motion, detailed textures, and lines that cross faces are harder and may leave visible traces, so review the full clip before publishing.
Upload one video per task. The upload field lists the current file requirements, and the highest-quality original you have will give the most convincing reconstruction.
Tasks use pay-as-you-go credits based on the duration of the source clip. The estimate beside the Generate button updates after you upload, so you can see the expected charge before you submit.
Upload a captioned clip, confirm the estimate, and get a clean version that is ready for new subtitles or a fresh edit.
Try it now
