Remove Text from Video
Erase any burned-in text from a clip — channel bugs, watermarks, lower-thirds, titles, and labels anywhere in the frame. Upload a video, let the AI rebuild what was underneath, and download it clean with the audio untouched.
Sign in to use this tool
To remove text from a video, upload a clip, choose subtitles or all on-screen text, and start the AI erase job. The result is a processed MP4 with the original audio and timing preserved.
- Formats
- MP4, MOV, WebM
- Maximum
- 10 minutes (can take hours)
- Pricing
- Credits by duration
- Audio
- Passed through
In short
To remove text from a video: sign in, upload your MP4, keep it on “all on-screen text,” and tap Remove text. The AI tracks every block of burned-in text across the frames, paints it out, and rebuilds the moving picture underneath, then re-encodes the clip with the audio intact. It clears corner bugs, watermarks, lower-thirds, titles, and labels — anywhere in the frame, not just the bottom. Charged by clip length, refunded if a job fails.
On-screen text lives all over the frame
Subtitles sit in a predictable strip along the bottom. The rest of a video’s text does not. A channel bug hides in a corner, a sponsor lower-third slides across the lower third, a title card fills the centre, a price tag or “NEW” sticker pops wherever the editor dropped it. Cleaning all of that means looking at the whole frame, not one band.
And like hardcoded captions, this text is baked into the pixels. There is no layer to hide and no menu to switch it off — the lettering is part of the picture, sitting on top of whatever was behind it. Removing it cleanly means rebuilding that hidden surface across every frame while the scene keeps moving.
That is what this tool does automatically: it finds each block of text, tracks it as the shot changes, erases the glyphs, and reconstructs the moving background underneath, then re-encodes the result. Static elements like a fixed corner watermark are the easiest of all, because the surrounding frames give the AI plenty to rebuild from.
How to remove text in 3 steps
- 1
Upload the video
Drag an MP4, MOV, or WebM onto the upload card, or tap to browse. Clips up to ten minutes upload straight from your browser to secure storage — files over 100 MB are compressed on your device first, and anything over 500 MB is refused — so a screen recording or a long export goes up without hitting a server limit. A progress bar tracks the transfer the whole way.
- 2
Switch to “All on-screen text”
This page opens in text mode, so the AI hunts for burned-in text anywhere in the frame — not just a caption strip along the bottom. Leave the region on “Whole frame” to catch corner bugs, top banners, and centred labels; or narrow it to a strip if every bit of text sits in one band.
- 3
Download the clean video
Tap Remove text. The AI tracks each block of text across every frame, paints it out, and rebuilds the moving picture underneath, then re-encodes the clip. Expect about 4 minutes of processing per 10 seconds of footage, so a full 10-minute clip can take a few hours. You don't have to keep the tab open: reopen this page in the same browser within 24 hours and the cleaned video plays right here with a Download button. The original audio, length, and timing are untouched.
Text this removes — and where to aim
| Text type | Typical example | Best region |
|---|---|---|
| Channel bugs & watermarks | A semi-transparent username or logo in the corner | Whole frame |
| Lower-thirds & name plates | Speaker names, titles, and graphic strips | Bottom strip |
| Title cards & headlines | Big intro text or a centred quote | Whole frame |
| Hardcoded captions | Burned-in subtitle lines you can't toggle | Bottom strip |
| Stickers & overlay labels | "NEW", price tags, app-added text overlays | Whole frame |
Real results, before and after
These are real jobs, not staged demos. Users of our sister site VideoTextRemover, which runs the same eraser as the tool above, uploaded these clips between October 19 and 23, 2026. We cut each pair to the same few seconds and dropped the audio. Every frame is as the eraser returned it, including the one that went wrong.
Product ad · talking head
Bold captions off a headset ad
Two-line white captions sat right over the presenter's hands and the headset he was holding. The user boxed the caption area, and the hands, the coiled cable and the jacket underneath came back intact. A presenter who stays in roughly one spot gives the eraser plenty of clean frames to rebuild from.
- Removed
- White word-by-word captions
- Mode
- Boxed region
- Shown
- 8 s of a 30 s clip
- Source
- Real user job, unedited output
Console ad · studio portrait
Two-line subtitles off a plain backdrop
Blue-and-white subtitles over a grey hoodie and a plain studio backdrop: small text, a quiet background and a fixed position, the three things that make removal clean. Subtitles-only mode cleared both lines and left no smudge on the hoodie.
- Removed
- Two-line English subtitles
- Mode
- Subtitles only
- Shown
- 9 s of a 37 s clip
- Source
- Real user job, unedited output
Fan video · handheld phone
A caption off a shaky red-carpet clip
Handheld footage, a moving crowd, and a beaded dress directly behind the caption: about the busiest background text can sit on. The user boxed the caption, and the beading and the crowd were rebuilt as the camera swung around.
- Removed
- One-line fan caption
- Mode
- Boxed region
- Shown
- 8 s of a 24 s clip
- Source
- Real user job, unedited output
Fitness clip · mirror selfie
A tiny corner tag off a dark wall
A small vertical Korean tag sat against the dark gym wall at the left edge. A small mark on a dark, mostly still background is the easiest job there is, and one box cleared it for the whole clip.
- Removed
- Small Korean text tag
- Mode
- Boxed region
- Shown
- 6.8 s, the whole clip
- Source
- Real user job, unedited output
Channel intro · where it went wrong
The tagline went, but so did part of the logo
The line under the title came out clean. But subtitles-only mode also read the small word inside the round logo as text and smeared it, while the larger lettering survived. When text you want to keep sits near text you want gone, draw a box around only the part to erase before you start.
- Removed
- Tagline under the title (and, by mistake, a word in the logo)
- Mode
- Subtitles only
- Shown
- 5.4 s, the whole clip
- Source
- Real user job, unedited output
What people use it for
Clean B-roll and stock-style footage
Strip channel bugs, usernames, and burned-in labels off clips you have the right to reuse, so the footage reads as neutral B-roll in your edit.
Reuse a render that shipped with text baked in
When the only copy of a clip already has titles or lower-thirds rendered into it, recover a text-free version to recut or relayout without the old graphics fighting the new ones.
Remove watermark text from your own exports
Tools and templates sometimes stamp a watermark into the output. Erase the burned-in text and keep the picture underneath intact.
Prep clips for re-captioning or translation
Clear the original on-screen text first, then add your own fresh titles or translated captions on a clean canvas.
Tips for cleaner results
Whole frame for scattered text, strip for a single band
If text appears in more than one place — a corner bug plus a bottom caption — keep the region on the whole frame. Only narrow to a strip when every piece of text genuinely lives in one horizontal band.
Static text over moving footage rebuilds cleanest
A fixed corner watermark over a moving scene is the easy case — the AI has plenty of surrounding frames to rebuild from. Text laid over fine, fast detail is harder and may leave faint residue on a few frames.
The output matches your source
The cleaned clip comes back at the source resolution (capped at 1080p) after a single re-encode — no quality settings to pick. If you need a lighter file for chat or email, compress the downloaded MP4 afterwards.
Trim first — you pay by length
Processing and credits scale with the clip's duration, so cut it down to just the part you need before uploading rather than processing footage you'll throw away.
Clearing a channel bug or a title card is often step one of reusing an old clip — and old clips tend to be soft or low resolution to begin with. This tool hands the video back at its source resolution (1080p at most), so if you need more, take the downloaded MP4 to our sister site VideoEnhancer: upscale the video to 2K or 4K, run the AI video enhancer on footage that looks muddy, or use video restoration to reduce grain in tape transfers and archive clips. Clips of up to two minutes work, with a five-second free trial.
What it can and can’t do
It erases text — it doesn't read or translate it
The AI rebuilds the picture under the text; it never decodes the words. To translate, remove the original text here, then add a new track in your editor.
Hidden detail is reconstructed, not recovered
Where text covered part of a face or object, the AI predicts a plausible continuation from the surrounding frames — seamless over plain or moving backgrounds, an educated guess over intricate detail.
Only clean footage you have the right to edit
Removing a watermark, credit, or channel bug from someone else's video can carry its own legal weight beyond the underlying copyright. Use it on your own content or material you're licensed to edit.
Want to see how it handles a title card or a timestamp in your clip before you buy credits? Our sister site VideoTextRemover runs the same on-screen text eraser on a site made just for video, and once you sign in with Google the first 30 seconds of video each month are free. Its accounts and credits are separate from MagicRemover's.
Questions about removing text from video
More MagicRemover tools
Remove Subtitles from Video
The sibling tool tuned for the bottom caption track — hardcoded subtitles you can't switch off.
Remove Watermark from Video
For logo and stamp marks text detection can't catch — box the watermark and the AI rebuilds what's behind it.
Remove Text from Image
The single-image version — erase captions, date stamps, and writing from one photo or screenshot.
Remove Watermark from Photo
Scrub diagonal © lines, stock banners, and logo overlays off still images you own.
Powered by apimodels.app — a unified AI image generation API