Remove text from video
Clean up a title, caption or label that is already part of your video picture. Upload an MP4 or MOV, choose a removal mode and review the result. If you still have the editing project, try removing the text layer and exporting a clean version first.
Choose your source and mode
Review the scope and price
Inspect the complete export
Quick answer
Choose Remove text to detect on-screen writing automatically. Choose Remove objects to describe a specific target when other writing should stay. Both methods reconstruct covered pixels, so inspect the complete result before using it.
Upload your source, review the quote and check which writing should stay before starting.
Automatic text removal is temporarily unavailable. No credits will be charged. You can switch to Remove objects and describe the text to use the current method.
Before uploading, turn off subtitles in your player or hide the text layer in your editing project. Watch the same moment again. File format alone does not settle the question: an MP4 can contain separate subtitle data and words already burned into its picture.
Scroll sideways to read all columns.
| What you find | Best starting point | Why |
|---|---|---|
| An editable title or text layer | Hide or delete the layer and export again. | Retain the source picture without estimating covered pixels. |
| A selectable subtitle stream or separate file | Disable or exclude that subtitle data. | The words are stored separately from the picture. |
| Words remain after those checks | Treat them as part of the picture. | Removing them requires replacing the affected pixels. |
The words you want to keep matter as much as the words you want to erase. Check the active mode before submitting.
Remove text needs no prompt or selection box. Use it when the writing across the clip is what you want cleaned up. It may also remove signs, product labels or logos you wanted to retain. A talking-head clip with captions and a branded product needs extra care.
Remove objects lets you describe one title or label in a short English sentence. Name its position, colour and appearance. This is a target description rather than a protection mask; compare the other writing after processing.
If automatic text removal is unavailable, the tool explains that and offers the object workflow. Read the current mode, duration and cost immediately before submitting. Availability does not guarantee an acceptable result for every scene.
Keep an untouched source so you can compare the result. For an initial evaluation, choose a continuous shot that includes the hardest moment rather than only a quiet opening frame.
Upload MP4 or MOV, up to 100 MB, 1080p and 60 fps. Start with the highest-quality source you have and avoid repeated compression. Use constant-frame-rate footage; convert variable-frame-rate video first. Choose automatic text detection or a specific object request.
The tool displays the available duration and quote. One credit covers one billable second, with partial seconds rounded up. Eligible verified accounts receive 3 trial credits; that does not cover an arbitrary full-length video. Confirm any scope change and the Keep original audio option before starting.
Play at normal speed, then pause where the wording changes or the camera moves. Check the first and last frames, the repaired texture and any subject beside the text. Listen to the exported audio and compare its timing with the picture before downloading.
These suggested descriptions apply to Remove objects. Automatic text mode needs no prompt. Replace the colour and location with what appears in your clip. “Remove text” is ambiguous when the frame contains a headline, captions and a sign.
Remove the white title in the upper left corner.Remove the yellow label above the door on the right.Remove the small text overlay along the bottom edge.Choose the method by looking at what sits behind the words. A plain wall is a different task from a moving face or a product serial number.
If the words are still an editable layer, remove that layer. This preserves the underlying footage without reconstructing the covered area.
Replaces the covered area instead of deliberately blurring a rectangle. It can retain the composition, but the rebuilt patch may look soft, flicker or contain altered detail. “Without blur” describes the method, not guaranteed sharpness.
Useful if the text sits in a margin you can discard. Cropping changes the composition and may cut out hands, products or other details you need.
Can be appropriate when hiding the words matters more than rebuilding a natural-looking picture. The cover remains visible; it does not recover the hidden background.
A single clean frame can hide a problem during motion. Compare normal-speed playback and difficult frames against the untouched source.
Scroll sideways to read all columns.
| Problem | Where to look | Next step |
|---|---|---|
| Remaining letters or a shadow | When the wording changes. | In object mode, clarify the target; evaluate a shorter section. |
| A flickering patch | Camera moves and moving textures. | Inspect adjacent frames and playback, not just a screenshot. |
| Distorted subject detail | Faces, hands or product edges beside the text. | Use a clean source or a controlled desktop repair if details matter. |
| Other writing disappeared | Signs, labels and logos you meant to keep. | Try a targeted object request instead of automatic detection. |
| Audio or timing seems wrong | Beginning, end and visible speech. | Compare source timing and check the audio option before another edit. |
Text covers image information. Reconstruction estimates a replacement; it cannot establish the true appearance of an obscured face, word or identifying detail. The illustration above shows the concept and is not a validated removal result.
Faces, fingers, patterned clothing, scrolling letters and cuts between shots can produce altered detail or visible artifacts. If those details are essential, prefer a clean source or an editor with more manual control.
Each processed file may introduce another round of image changes and encoding. Return to the original source, identify the cause and change the approach before spending more credits on another attempt.
Distinguish optional subtitle data from words already rendered into the picture.
Use position, colour and appearance to make an object request more precise.
Review processing costs and retention before choosing a longer edit.
Use object mode to describe one target and compare everything you wanted to keep. Automatic text mode has no local selection box and may remove other writing or logos.
Text mode processes the selected video within the duration displayed in the tool. It does not silently shorten the video. If credits are insufficient, review an explicit scope choice or add credits before submission.
Keep original audio is selected by default. Check the option before submitting and listen to the entire exported file afterward.
If you have the editing project, delete or hide the separate text layer and export again. If the words are already part of the picture, they need picture repair. Deleting the video clip is a different operation.
Upload your source, review the quote and check which writing should stay before starting.