Why Removing Burned In Subtitles Is Hard
Subtitle removal is one of the most challenging video editing tasks. Burned in (hardcoded) subtitles are part of the video's pixel data they cannot simply be toggled off. The challenge is bigger than removing a static logo because subtitles:
- Change every few seconds: Each line is different, with different lengths, line breaks, and durations. Hundreds of unique text shapes per video.
- Sit on the lower third of the frame: Often overlapping with faces, hands, body movement, and important visual details.
- Have a "safe area" background: Subtitle creators often add a black or semi transparent box behind text for legibility, which has to be removed along with the text itself.
- Use varied fonts and colors: White with black outline is most common, but yellow, fancy fonts, and colored highlights all exist.
Wipe AI's auto detect mode handles all of these because it identifies text regions on a frame by frame basis, then inpaints each region using surrounding context.
When Subtitles Can Be Removed Without AI
Before using AI inpainting, check whether your video has soft subtitles (a separate text track) instead of hardcoded ones. To check:
- Open the video in VLC Media Player.
- Click Subtitle → Subtitle Track.
- If you see options other than "Disable", they're soft. Select "Disable" the subtitles disappear instantly with no AI processing required.
- If subtitles remain visible after disabling all tracks, they're hardcoded.
Common sources of soft subtitles: Netflix downloads, official streaming releases, YouTube videos with captions enabled, MKV files with embedded SRT tracks. Common sources of hardcoded subtitles: foreign films, anime fansubs, social media reposts, TikTok auto captions, news broadcasts, and pirated/redistributed content where the original subtitle track was lost.
Removing Subtitles in Different Languages
Wipe AI's auto detect works on multiple writing systems with no manual configuration:
- Latin scripts: English, Spanish, Portuguese, French, German, Italian, Turkish, Polish, Dutch, Romanian detected reliably.
- Cyrillic: Russian, Ukrainian, Bulgarian, Serbian detected with similar accuracy.
- CJK: Chinese (Simplified and Traditional), Japanese, Korean detected, though dense character text can sometimes leave faint outlines.
- Arabic, Hebrew, Persian: Right to left scripts work, with the same caveat about dense glyphs.
- Devanagari (Hindi): Detected, though the headline strokes (shirorekha) sometimes confuse the OCR boundary detection. Manual selection often gives better results.
- Thai: Detection works, but tonal marks above letters can be missed if they're outside the main text bounding box.
Karaoke Lyrics and Animated Subtitle Removal
Karaoke style subtitles where each word lights up as it's sung, often with bouncing cursors and color animations are technically the same as regular hardcoded subtitles to the AI. The animation doesn't change the pixel removal task; the AI processes each frame independently regardless of whether the text is animating.
That said, karaoke text is usually centered low in the frame and stays for the entire song, which means a lot of the video has to be inpainted. Processing time scales with the area covered: a 3 minute karaoke video might take 4 to 6 minutes to clean.
Subtitle Removal vs Translation Workflow
Some users want to remove existing subtitles in one language and add their own subtitles in another language. The recommended workflow:
- Strip the original subtitles with Wipe AI's auto detect mode.
- Run the cleaned video through speech to text (Whisper, Google Speech, or any captioning tool) to generate fresh subtitles in your target language.
- Burn the new subtitles in using FFmpeg, HandBrake, or any video editor.
This produces a clean, translated video without traces of the original subtitles.
Common Subtitle Removal Mistakes
- Cropping the bottom third: Removes subtitles but also cuts off important video content. Not viable for vertical TikTok style videos.
- Black bar masking: Hides subtitles behind a black rectangle. Looks worse than the original subtitles. Not a real solution.
- Manual frame by frame editing: Theoretically possible in DaVinci Resolve or After Effects, but takes hours per minute of video. Not practical at any scale.
- Re encoding without removing: Some users hope re encoding will "wipe" subtitles. It will not re encoding preserves whatever pixels are in the source.
For broader text removal beyond just subtitles, see remove text from video. For platform specific watermarks, see TikTok Watermark Remover, CapCut Watermark Remover, and Kinemaster tools.