Remove burned-in subtitles

Erase dialogue already rendered into your footage and reconstruct the picture behind the words with AI.

Drop your video here

or

What does “burned-in” mean?

Burned-in subtitles are letters rendered into the image itself. People also call them hardcoded subtitles, baked-in text, or embedded subtitles, although “embedded” can sometimes refer to a separate stream inside a container. In the first sense, the words occupy picture pixels and cannot be disabled with a player setting.

Start by turning text display off in your player. Next, inspect the editing timeline for a separately selectable layer. Finally, check for an extractable .srt or .vtt file or stream. If the words stay visible after every text layer is disabled, the visible version is part of the picture. A clip may contain both rendered words and a separate file, so finding one does not prove that the picture is clean.

If a selectable layer is the only source, hide it or exclude that stream while remuxing. Keep the original file and any sidecar text.

Compare three ways words reach the screen

Words rendered into video

Burned in subtitles and hardcoded subtitles describe this fixed picture layer. A player cannot hide those pixels; changing them requires a visual edit such as AI reconstruction.

Selectable text

Soft subtitles are stored separately, either as a file or as a subtitle track inside the container. Turn them off or remove the stream without altering image frames.

Always-visible access text

Open captions may include speaker names and sound descriptions rendered into the picture. They also cannot be toggled off and can be candidates for the same visual process.

Why hardcoded words cannot be switched off

A finished export no longer stores the original background underneath each glyph as a separate layer. The processing model identifies the targeted text over successive frames and estimates what belongs there from surrounding visual information. It is an approximation, not a guarantee that lost detail can be recovered exactly.

Cropping would discard part of every frame, while masking or blurring would hide the same portion behind a patch. Reconstruction aims to preserve the full composition instead. Busy patterns, camera movement, and faces passing behind letters can still make the outcome less convincing; check the preview before using the result in a final edit.

Remove burned in subtitles in three steps

  1. 1

    Pick a source file

    Choose a supported clip and sign in. Use automatic detection or mark a region yourself, then choose the original size or an available smaller output.

  2. 2

    Process the picture

    Submit the selection for frame-by-frame analysis and background reconstruction. Time and credit usage depend on the clip length, selected mode, and output resolution.

  3. 3

    Review the export

    Watch the processed clip and download an MP4 without an added watermark. No installation is needed; the video is re-encoded, so quality or frame rate can change.

What to expect from detection

The model is designed around subtitle-style lettering, not every object in a scene.

Handled well

  • Dialogue lines in the lower or central portion of a frame
  • Translated text rendered over the original footage
  • Lines that change as speakers continue a conversation
  • Subtitle-style writing in multiple scripts and orientations

Usually left alone

  • Corner watermarks, channel logos, and platform badges
  • Scrolling banners along the top of the frame
  • Scene text on signs, screens, or packaging
  • Separate text streams, which do not need image reconstruction

Out-of-area elements are not the intended target. Automatic detection can still miss words or affect nearby details, so review each result.

Which page fits your file?

For a selectable stream or an extractable text file, read the subtitle format guide before uploading. For on-screen speaker labels or sound descriptions, see the caption editing guide. If you have another overlay resembling dialogue, the main video text workspace provides the same modes.

Use the main text tool for a clip with hardcoded subs. Save an untouched copy and compare the preview.

Questions before you process

Do burned-in and hardcoded mean the same thing?

In this context both describe words rendered into the video image. They are not a selectable text layer, even if another track is also present.

How can I tell whether the words are a separate stream?

Disable the text option in your player and inspect the available tracks or your editing timeline. If the same words remain in the picture, that version is embedded; a video can contain both types.

Will every trace of the text disappear?

Not always. Fine patterns, moving objects, and complex edges can leave artifacts because the hidden pixels must be estimated; preview the result before using it.

Will a separate track survive, and does processing affect quality?

The exported MP4 is not guaranteed to retain separate tracks, so keep the source and any text files. Re-encoding can change quality or frame rate even when you keep the original resolution.

Which formats can I use, and do I need an account?

The tool accepts MP4, MOV, M4V, WebM, MKV, AVI, FLV, WMV, and TS; the output is MP4. Sign in with Google or email to process a clip, and new accounts receive 15 free credits.

Ready to inspect your footage?

Choose the source, process it, and check the result before downloading.

Choose a clip