About Video Subtitle Remover

Who we are

Video Subtitle Remover is a small, focused team building one thing well: tools that remove burned-in subtitles, captions and text overlays from video. The project started from a simple frustration shared by editors, translators and archivists — footage that is perfect except for text baked into the pixels, and no honest tool that fixes exactly that problem without cropping, blurring half the frame, or quietly re-encoding your audio out of sync.

We are not a general video editor, not an AI lab demo, and not a re-skin of someone else's pipeline. The site you are reading is built around a single processing contract: you confirm the region, the engine repairs the pixels inside it, and everything else about your file — audio, duration, frame order — passes through untouched. Every page, every FAQ answer and every limitation we publish is written to keep that contract explainable.

What we build

The product has three parts, all sharing one engine:

• Fourteen focused tool pages. Each targets one real search task — remove text from video, subtitle remover, caption remover, AI subtitle remover, remove subtitles from video online free, and their siblings. One keyword, one page, one default processing mode. The pages differ where the tasks genuinely differ: manual region selection for arbitrary text, bottom-band detection for subtitles, position-flexible masks for captions, and OCR-assisted proposals for the AI pages.
• The online tools. Short clips (up to 60 seconds, 720p, 100 MB per file) processed in the cloud during the free beta, with a small daily quota per person. Uploads and results are automatically deleted within 24 hours, and nothing is ever used for model training.
• The Windows application. The same engine packaged for Windows 10 and 11, with no length cap, batch folder queues, NVIDIA GPU acceleration with a CPU fallback, and zero uploads. Files processed in the desktop app never leave your disk.

The interactive demo at the top of every tool page is the real selection workflow — drag the box, auto-detect the subtitle zone, compare before and after — running entirely in your browser.

How the engine works

There are exactly two routes, and the tool always probes your file before choosing:

1. Track stripping (lossless). If the text lives on a switchable subtitle track — SRT, ASS and similar streams inside an MP4 or MKV — the engine drops the stream and copies everything else byte-for-byte. No re-encoding, no GPU time, finished in seconds. If your player can already toggle the text off, this is the only processing you need.
2. Region repair (inpainting). If the text is burned into the pixels, you confirm a rectangle over it. The engine rebuilds that region on every frame from the surrounding pixels and neighboring frames in time, then muxes the original audio back into the export. The mask is yours to adjust before anything runs, and the before/after preview is the checkpoint we never skip.

This two-route design is why we publish pages explaining when you should not use the repair pipeline at all. A tool that runs the GPU on a file that only needed a remux is wasting your time and quality.

The principles we publish

• Honest previews over success badges. Repair quality depends on what sits behind the text. Calm gradients rebuild almost invisibly; fast motion softens; faces under captions come back plausible, not perfect. We say so on every page, and the preview exists so you can judge before exporting.
• Original audio, always. Both routes preserve the audio stream, duration and frame order. A removal tool that breaks sync is worse than the text it removed.
• Privacy with numbers, not adjectives. Online: private storage, signed download links, automatic deletion within 24 hours, no training on uploads. Desktop: nothing uploads at all. The privacy policy spells out every category in plain language.
• No fake functionality. While the online beta is not yet open, the pages say exactly that — there is no upload box pretending to process. The Windows beta downloads publish on the Windows page with checksums and a version history.
• Rights stay yours to respect. Removing a subtitle or watermark transfers no rights. Every page carries the reminder to edit only footage you own or are licensed to modify, and DRM-protected files are refused outright.

Six languages, written not translated

The site ships in English (default), French, German, Spanish, Japanese and Korean. Each language has its own keyword research — the French pages target the phrasings French users actually type, the Japanese pages target Japanese search patterns, and so on. The copy is written natively in each language rather than machine-marched through a translation memory, because a page that reads like a translation neither ranks nor convinces.

The free tier, in numbers

• 60 seconds maximum clip length online
• 720p maximum online output resolution
• 100 MB maximum upload size
• 24 hours maximum retention of uploads and results
• 5 regions selectable per clip
• 6 languages across 15 tool pages — that is 90 localized pages, plus privacy and terms in each language

The Windows app removes the length, size and quota limits because it runs on your hardware. Both the online beta and the desktop beta are free; if a paid tier ever appears, the beta feature set stays free — that commitment is written into the terms of service.

Where the project is going

The roadmap is short and public: open the online processing queue, publish the Windows installer with its SHA-256 checksum, and keep expanding the test matrix (codecs, scripts, stylized fonts) with results reported honestly on the Windows page. We would rather ship a smaller feature list that behaves exactly as described than a long one held together by asterisks.

If you want to know how your files are handled before you upload the first one, read the privacy policy — it was written to be read, not skimmed past a cookie banner. And if something on this site claims more than it delivers, the contact page is the fastest way to tell us.