Hardcoded subtitles are part of the picture, so turning off captions will not remove them. The practical option is to mark the area where the text appears and reconstruct that part of each frame.
The quality of the result depends as much on the selection as the model. A tight box over a steady subtitle band is easier to repair than a large box that includes faces, hands, or moving objects.
Key takeaways
- Confirm that the subtitles are burned into the picture.
- Keep each removal box as tight as possible.
- Use separate boxes for captions and corner logos.
- Review motion and scene changes before downloading.
1. Check whether the subtitles are hardcoded
Try disabling captions in the player. If the text remains, it is burned into the video. If it disappears, export or disable the subtitle track instead; erasing pixels is unnecessary.
2. Mark the subtitle area
- 1
Upload the original
Open the AI eraser and choose a video up to 500MB.
- 2
Choose subtitle removal
A starting box appears near the lower part of the frame.
- 3
Tighten the box
Drag its sides and corners so it covers the text without taking in more background than needed.
- 4
Add another region if needed
Use a separate box for a second caption position or fixed logo. Each type supports up to five regions.
3. Review the places most likely to show artifacts
Check shots where people cross the subtitle band, the camera moves quickly, or the background contains rails, patterns, and small text. Pause on those moments and compare the same frame before and after.
If the repaired area looks soft, reduce the box height and submit another pass before changing anything else.
When erasing is not the best choice
- The subtitle covers a face for most of the shot.
- Large animated titles move across the whole frame.
- The background contains detailed text that must remain exact.
- You do not own or have permission to edit the source.
Frequently asked questions
Can hardcoded subtitles be turned off?
No. They are pixels in the video. They need to be removed and the covered area reconstructed.
How many subtitle areas can I mark?
Vimorph currently supports up to five regions of each removal type.
Why does the result sometimes look soft?
The selected area may be large or contain fast motion and detailed texture. A tighter selection usually gives the model more usable context.