How to Remove Hardcoded Subtitles From a Video (2026)

First check whether the subtitles are soft (a separate track you can switch off) or hardcoded (burned into the picture). Soft subtitles come out losslessly in seconds: ffmpeg -i input.mkv -map 0 -map -0:s -c copy output.mkv. Hardcoded subtitles are pixels, so you crop them off, blur or cover them, or use AI inpainting, which detects the text and fills the background behind it.
People usually want this for one of three reasons: the old subtitles are in the wrong language, they have typos, or they clash with new captions. Whatever the reason, only do it on video you own or have the rights to edit; more on that below.
Facts checked October 20, 2026. FFmpeg commands follow the official FFmpeg documentation and were run against FFmpeg 9.0.1 test files to confirm the syntax. Tool limits come from each vendor's own pages (links under Sources); our AI video inpainting roundup covers several of these tools in more depth.
Are your subtitles soft or hardcoded?
Soft subtitles are stored as a separate track inside the file (or as a sidecar file such as SRT), so a player can show or hide them. Hardcoded subtitles, also called burned-in or open subtitles, are part of the video image itself. Our guide to captions vs subtitles vs SDH covers the open and closed distinction in more depth.
| Soft subtitles | Hardcoded (burned-in) subtitles | |
|---|---|---|
| Where they live | A subtitle stream in the container, or a sidecar file | Pixels in every frame |
| Can the viewer turn them off? | Yes, in a compatible player | No |
| How to remove | Drop the track (remux); no quality loss | Crop, blur or cover, or inpaint; the picture is altered |
| Time needed | Seconds | Minutes to hours, depending on method |
Two quick tests:
- Look for a subtitle toggle. If your player's subtitle or CC menu turns the text off, it's a soft track.
- List the streams.
ffprobe -v error -show_entries stream=index,codec_type,codec_name -of compact input.mkvprints every stream. A line withcodec_type=subtitle(for examplesubripormov_text) is a soft track. If there's no subtitle stream but text still appears on screen, it's burned in.
How do you remove soft subtitles without re-encoding?
Remux the file without its subtitle streams. With stream copy (-c copy), FFmpeg's documentation notes, "there is no decoding or encoding, it is very fast and there is no quality loss."
ffmpeg -i input.mkv -map 0 -map -0:s -c copy output.mkv
-map 0 selects every stream in the input and -map -0:s is a negative mapping that removes the subtitle streams; FFmpeg's documentation uses the same pattern to "map all the streams except" a given one. The same command works for MP4 files with mov_text subtitles.
You'll also see the shorter ffmpeg -i input.mkv -c copy -sn output.mkv. As an output option, FFmpeg says -sn "disables subtitle recording i.e. automatic selection or mapping of any subtitle stream." The catch: without -map, FFmpeg's automatic selection keeps only one video and one audio stream, so a file with several audio languages loses all but one. In our test file with two audio tracks, -sn kept one; the -map 0 -map -0:s version kept both.
To drop just one subtitle track and keep the others, target it by index: -map 0 -map -0:s:0 -c copy removes the first subtitle stream only.
If you prefer an app, MKVToolNix's mkvmerge drops every subtitle track with -S (--no-subtitles). HandBrake lets you pick which subtitle tracks to keep from the "Tracks" button on its Subtitles tab, but it re-encodes the video, so remuxing is faster and lossless.
How do you remove hardcoded subtitles?
Burned-in text can't be switched off, so every method changes the picture. There are three, from simplest to most involved.
Method 1: Crop the subtitle band
If the subtitles sit in a strip at the bottom and nothing important happens there, crop it away. This removes 140 pixels from the bottom of a 1080p video:
ffmpeg -i input.mp4 -vf "crop=iw:ih-140:0:0" -c:a copy output.mp4
The result is no longer 16:9. To restore 1920 × 1080, scale up and trim the sides, which zooms the image by about 15%:
ffmpeg -i input.mp4 -vf "crop=iw:ih-140:0:0,scale=-2:1080,crop=1920:1080" -c:a copy output.mp4
Filters like crop re-encode the video (stream copy can't be combined with filters), so add -crf 18 if you want higher quality than FFmpeg's default. Cropping is clean and fast, but you lose part of the frame, and it doesn't work when subtitles move or sit over the action. Our aspect ratio guide helps if you're reframing for another platform anyway.
Method 2: Blur or cover the text
Blurring or covering hides the text without pretending it was never there. It's honest and quick, and it suits a band where you'll place new captions on top. To blur a 120-pixel band that starts 160 pixels above the bottom:
ffmpeg -i input.mp4 -filter_complex "[0:v]crop=iw:120:0:ih-160,boxblur=20[b];[0:v][b]overlay=0:H-160" -c:a copy output.mp4
FFmpeg's delogo filter is another option. It "suppress[es] a TV station logo by a simple interpolation of the surrounding pixels": -vf "delogo=x=100:y=930:w=1720:h=110". Add :show=1 to draw a green box while you line it up. In our test, FFmpeg 9 refused a box that touched the frame edge ("Logo area is outside of the frame"), so keep the rectangle inside the picture. FFmpeg's own docs warn that "sometimes something even uglier" appears, and a wide subtitle band gives the interpolation a lot to fill.
Method 3: Use AI inpainting with automatic text detection
AI inpainting finds the text, masks it in every frame and generates the background behind it. Some tools detect text automatically; others need you to brush or box the area once and then track it. It's the only method that can leave a full, clean frame, and also the one most likely to show artifacts. For how inpainting works, see our inpainting roundup.
Tools that offer it, as described on their own pages in October 2026:
| Tool | How it finds the text | Documented limits | Free option |
|---|---|---|---|
| video-subtitle-remover (open source) | Automatic detection of all text, or an area you set | Runs locally with Python; uses CUDA, Apple Silicon, DirectML or CPU | Free (Apache-2.0) |
| Vmake | Subtitle mode deletes text automatically; you can set removal and protected areas | Limits not stated | Not stated |
| Unwatermark | "Detects text frame by frame" | MP4, MOV or M4V under 500 MB | Free version processes the first 6 seconds |
| Media.io | Brush, text prompt (AI Model) or Gaussian Blur | MP4 or MOV up to 15 minutes | 3 daily credits; 720p with watermark |
| Filmora AI Video Object Remover | Magic Box or Smart Brush; Filmora suggests box mode for subtitles | Desktop editor | Free version watermarks exports |
| HitPaw Watermark Remover | Five region modes, from AI Mode to Gaussian Blur | Windows and Mac | Free trial watermarks exports |
| Runway Remove from Video | Text prompt, no mask | Clips up to 5 seconds; 15 credits per second; Standard plan or above | No |
| After Effects Content-Aware Fill | A loose mask around the text | Desktop; needs compositing skill | 7-day trial |
| DaVinci Resolve Studio Object Removal | Power Window, tracking, then scene analysis | Studio only (US$295) | No |
The open-source video-subtitle-remover is the most direct match for "automatic text detection": its README says it supports "automatic removal of all text throughout the entire video (without inputting a position)," with several inpainting models, from a fast one suited to live action to ProPainter, which "consumes a significant amount of VRAM." It needs a local setup. Runway's 5-second limit means a long video has to be split into pieces, and Unwatermark's free version stops after 6 seconds. CapCut, covered in our roundup, also lists a subtitle-removal page, but its site blocked our checker this time; confirm in the app.
Which method should you use?
| Method | Picture after removal | Speed | Cost | Best when |
|---|---|---|---|---|
| Remux (soft subtitles) | Untouched | Seconds | Free | The subtitles are a separate track |
| Crop | Clean, but part of the frame is lost | Fast | Free | Text sits in a band with nothing important |
| Blur or cover | Visible blur or box | Fast | Free | You'll place new captions over the same area |
| AI inpainting | Full frame, possible smearing or flicker | Slow to render | Free to paid | Simple backgrounds and short clips |
| Re-export from the source | Perfect | Depends on project | Free | You still have the original project or clean master |
The last row is the one people forget. If you made the video, re-exporting from the original project without the subtitle layer beats any removal method.
Why does subtitle removal fail on some videos?
Inpainting has to invent what was behind the text, and some footage gives it little to work with:
- Busy or moving backgrounds. Crowds, foliage, water and handheld camera moves make the fill swim or flicker from frame to frame.
- Text over faces or detailed objects. The model can't know what a mouth or a logo looked like under the letters.
- Large, outlined or boxed subtitles. Thick outlines, drop shadows and background boxes cover more area to rebuild.
- Moving or animated text. Karaoke-style captions and text that changes position each line are harder to detect and track.
- Heavy compression. Blocky, low-bitrate video leaves soft halos around letters that detection can miss.
Check the result frame by frame on the hardest shots, not just at normal speed.
Is it legal to remove subtitles from a video?
Removing subtitles is a technical edit, not a license. Do it on content you created, commissioned or licensed with the right to modify it. Stripping burned-in text from someone else's film or show to re-upload it is still copyright infringement, and YouTube's monetization policy on reused content, which covers repurposing content without significant original commentary or modification, "applies even if you have permission from the original creator." In the US, knowingly removing copyright management information, such as a title, author or copyright notice, to enable infringement can also violate 17 U.S.C. §1202. Ordinary dialogue subtitles aren't on that statutory list, but burned-in credits or a channel watermark can be. This isn't legal advice.
Watermarks are a separate question; see video watermarks explained for the rights side.
What should you do after removing the old subtitles?
Most people remove subtitles to replace them. Keep the clean video as your master, and keep new captions as a separate file so you never have to do this again. Our SRT vs VTT guide explains which format each platform takes, and the free SRT converter converts between them. If you're replacing subtitles with another language, compare subtitles and AI dubbing.
Loopdesk (disclosure: Loopdesk is our product) doesn't remove burned-in text. Where it helps is the next step: its AI assistant, Aura, transcribes your video and generates captions in 108 languages, including mixed-language and right-to-left scripts, in the browser. There's no free plan; Creator Lite is $19 a month.
Sources
Read on October 20, 2026.
- FFmpeg documentation (
-map,-sn,-c): https://ffmpeg.org/ffmpeg.html - FFmpeg filters (
crop,boxblur,overlay,delogo): https://ffmpeg.org/ffmpeg-filters.html - FFprobe documentation: https://ffmpeg.org/ffprobe.html
- MKVToolNix mkvmerge documentation: https://mkvtoolnix.download/doc/mkvmerge.html
- HandBrake documentation, Subtitles: https://handbrake.fr/docs/en/latest/advanced/subtitles.html
- video-subtitle-remover (GitHub): https://github.com/YaoFANGUK/video-subtitle-remover
- Vmake: https://vmake.ai/remove-subtitles-from-video · Unwatermark: https://unwatermark.ai/video-subtitles-remover/
- Media.io: https://www.media.io/video-watermark-remover.html · https://www.media.io/pricing.html
- Filmora AI Video Object Remover: https://filmora.wondershare.com/ai-video-object-remover.html
- HitPaw Watermark Remover: https://www.hitpaw.com/remove-watermark.html
- Runway Remove from Video: https://runway.com/apps/remove-from-video
- Adobe, Content-Aware Fill in After Effects: https://blog.adobe.com/en/publish/2019/04/02/explore-new-content-aware-fill-in-after-effects
- Blackmagic Design, DaVinci Resolve Studio: https://www.blackmagicdesign.com/products/davinciresolve/studio
- YouTube monetization policies (reused content): https://support.google.com/youtube/answer/1311392
- 17 U.S.C. §1202: https://www.law.cornell.edu/uscode/text/17/1202
Frequently Asked Questions
Can you remove hardcoded subtitles from a video?
Yes, but not perfectly. Burned-in subtitles are part of the picture, so you crop the band off, blur or cover it, or use AI inpainting to rebuild the background. Inpainting gives the cleanest full frame on simple backgrounds and can smear or flicker on busy, moving shots.
How do I remove subtitles from an MKV or MP4 file?
If they are a soft track, remux without it: ffmpeg -i input.mkv -map 0 -map -0:s -c copy output.mkv. This keeps every video and audio stream with no quality loss. For MKV files you can also use mkvmerge with -S. Burned-in text needs cropping, blurring or inpainting instead.
Is there an AI that removes subtitles automatically?
Yes. The open-source video-subtitle-remover detects and removes on-screen text without you marking its position, and web tools such as Vmake and Unwatermark offer automatic subtitle modes. Media.io, Filmora, HitPaw and Runway work from a brush, box or prompt. Results depend on the background.
Can HandBrake remove burned-in subtitles?
No. HandBrake can add or remove soft subtitle tracks and choose which to keep on its Subtitles tab, but burned-in subtitles are part of the image. It can crop the frame, which removes a bottom subtitle band along with that part of the picture.
Is it legal to remove subtitles from a video?
On video you own or have the rights to edit, yes. Removing subtitles doesn't give you rights to someone else's content, and re-uploading it is still infringement. In the US, knowingly removing credits or copyright notices to enable infringement can also violate 17 U.S.C. §1202. Not legal advice.
Why does the area look smeared after removing subtitles?
The tool has to invent what was behind the text. Busy or moving backgrounds, faces behind the letters, thick outlines and heavy compression all leave it little to work with. Try a tool with a stronger model, a tighter mask, or crop or blur instead.
Removed the old subtitles? Try Loopdesk - generate new captions in 108 languages from your browser and review every line before you export.