AI Subtitle Removal
Automatic detection
Detect and erase hardcoded subtitles automatically with AI, while the original audio track stays untouched.
Precise video cleanup
Auto-detect hardcoded subtitles, or select fixed watermarks and logos. Start with a free watermarked preview; clean downloads use credits.
Verified output


Try your own video
Choose your video and a removal method. We show whether it qualifies for a free preview or needs credits before processing.
Honest proof
Choose subtitles, a fixed watermark, a synthetic logo or a timestamp. Compare the verified stills, then play the 10-second original and processed sample without starting a new AI job.
Separate five-minute gates passed for H.264 MP4 landscape and H.264 MOV portrait inputs in both automatic-subtitle and fixed-box modes. They do not prove HEVC-encoded videos or near-full-width translucent subtitle backgrounds, which remain unsupported.
Automatic removal of hardcoded subtitles; this does not remove a separate caption track.


Sample source / license / changes
Source: Big Buck Bunny — copyright 2008 Blender Foundation / bigbuckbunny.org
Open source · License: CC BY 3.0
Change: Trimmed 10-second clip, resized to 960x540, and re-encoded for cached automatic subtitle-removal playback.
Duration 10s
One tool for every kind of mark
AI subtitle detection, precise box selection and clean output — everything is built around getting your video mark-free.
Automatic detection
Detect and erase hardcoded subtitles automatically with AI, while the original audio track stays untouched.
Box up to 2 areas
Draw tight boxes around one or two fixed watermarks, logos or channel badges. Each selected region is reconstructed frame by frame.
The swipe interaction
Precise region selection for a custom logo or badge that stays in one area. Boxes do not track marks that move around the frame.
Up to 60 seconds
For supported MP4 clips up to 60 seconds and 100 MB, use your free-preview allowance to inspect a watermarked 720p result. Clean downloads use credits; other supported inputs require prepaid credits.
Paid clips up to 5 minutes
Buy a one-time Quick Pack, use its credits across clean H.264 MP4 or MOV jobs up to 500MB, and download at up to the processed 1080p source size without a Markswipe watermark. Markswipe never upscales.
Auto deletion
Uploads are sent to our processing provider. Preview links work until their displayed expiry; clean downloads are private. Files are deleted through scheduled cleanup.
Clean up a video in three simple steps
Choose a video, review its processing terms, and remove the marks.
Start ErasingUpload a free MP4 up to 100 MB, or sign in for a prepaid H.264 MP4 or MOV up to 500 MB. Free previews accept clips up to 60 seconds; prepaid jobs can process one clip up to 5 minutes and 1080p.
Let AI detect hardcoded subtitles, or draw up to 2 fixed boxes around watermarks, logos and overlay text.
For eligible free MP4 clips, inspect the watermarked preview before you unlock. Any prepaid input—longer clips, MOV or files over 100 MB—produces an entitled clean download after processing.
Use your allowance across multiple clips. Credits remain valid for 12 months. Each supported paid clip can be up to 5 minutes. Free previews include a Markswipe watermark.
The credit pack is a total allowance across clips, not the maximum length of one clip. Current price and checkout terms are shown on the pricing page.
Upload an H.264 MP4 or MOV and remove fixed watermarks, logos or text — or automatically detect hardcoded subtitles. Review an eligible free MP4 preview or use prepaid processing for MOV, larger files and clips up to 5 minutes.
Everything you need to know about Markswipe
You can process an MP4 up to 60 seconds and 100 MB and inspect a watermarked 720p preview without paying. A free account saves tasks to My videos; it does not include clean downloads. Credits unlock clean downloads for H.264 MP4 or MOV jobs up to 5 minutes per clip, 500 MB and 1080p without upscaling.
Short clips usually take about 2–7 minutes. In the verified five-minute gates, jobs took about 5–16 minutes. Footage complexity and current service load can change the wait, and signed-in tasks remain available in Recent tasks.
Yes. The audio track is kept untouched — only the visual marks you select are erased.
Free direct uploads accept MP4 up to 100 MB and 60 seconds. Prepaid processing accepts H.264/AVC MP4 or MOV up to 500 MB, 300 seconds and 1080p. MOV/HEVC, WebM and inputs above 1080p are not supported, and outputs are never upscaled.
Free results are rendered at up to 720p with a Markswipe watermark. Credits unlock the clean download at the processed source resolution without the Markswipe watermark.
Uploads go to our third-party processing provider and are removed after processing reaches a final state, with cleanup retries after interruptions. New free previews remain available for 7 days; clean files remain available for 14 days after unlock or prepaid completion. The result page shows the actual expiry date. Preview links can be opened by anyone who has the link; clean downloads remain private. Only upload videos you have the right to process.
Yes, when it stays in a fixed region. Draw a tight box around a custom logo, badge or overlay that automatic detection does not recognize. Box selection processes the same coordinates across the clip; it does not track a mark that moves around the frame.
Yes — select up to 2 areas per video in the eraser studio, and they are all cleaned in a single pass.
The public before/after showcase includes verified hardcoded-subtitle, fixed watermark, synthetic logo and changing timestamp examples, with audio and duration preserved. The boxed examples use fixed regions and do not claim moving-mark tracking or automatic recognition of every brand. These are internal validation samples, not customer case studies; your own preview remains the evidence for your footage.
A small fixed mark over a flat or slowly changing background is the strongest case. Detailed textures, rapid motion, scene cuts or subjects passing behind the selected region can leave a subtle reconstructed patch. Automatic subtitle removal can also leave a visible band when subtitles sit on a near-full-width translucent background. Markswipe does not promise moving-object tracking or exact recovery of hidden pixels.