Picture quality guide

화면 자르기 없이 자막 제거

Updated · Editorial methodology

Keep the edges of your composition while repairing burned-in subtitles. This is useful when cropping would cut off a face, hands or other important detail.

Keeping the frame avoids losing picture at the border. It does not guarantee unchanged pixels: automatic repair estimates the region covered by subtitles, while a crop removes the entire strip.

Compare the repaired picture

Compare the original and result as the video plays.

원본
처리 결과
Actual AI result.

The cost of removing the subtitle band

In the dialogue sample, the subtitle shares its horizontal band with the cup and hands. Removing the bottom fifth of the image removes all of that band, not just the words.

Original frame

Original frame

The subtitle line crosses clothing and the area near the hands. The full 16:9 composition remains visible.

의도적인 흐림 효과

의도적인 흐림 효과

The letters become a visible softened band. It also softens the picture inside the selected rectangle.

Bottom 20% cropped

Bottom 20% cropped

The subtitle is outside the remaining frame, but part of the cup, hands and lower composition is lost. The image is not stretched to restore 16:9.

Automatic repair result

Automatic repair result

The supplied output has no subtitle at 1.00s and keeps the full frame. The subjects’ poses differ from the input sample, so do not use this still pair to score recovery of hidden detail.

The original, blur and crop use the input at 1.00s. Blur applies a 12-pixel box blur to a 680 × 90 source-pixel region; cropping removes the bottom 144 pixels. The supplied repair result is also sampled at 1.00s, but the subjects’ poses differ. It shows a text-free output and retained framing; this pair cannot establish pixel-for-pixel reconstruction quality.
하드코딩 자막 제거

Inspect the retained detail

A repaired output can keep the framing while changing detail where text used to be.

At 1.00s, the repaired version no longer shows the subtitle line. The hands, cup and sweater remain inside the frame, unlike the cropped demonstration.

원본 · 1.00s원본 video detail at 1.00 seconds
처리 결과 · 1.00s처리 결과 video detail at 1.00 seconds
The same crop coordinates show the lower band in each supplied video at 1.00s. Subject positions differ between the files; these are not pixel-aligned frames.

This pair shows that the output retains a full frame, but the differing poses prevent a controlled detail-restoration comparison. Inspect the moving output and its continuity before using it.

블러 없이 영상 글씨 지우기

Distinguish framing from processing duration

Two different changes are often described as “cutting the video.”

Framing and duration are separate
ChangeWhat becomes smallerWhat to check
Crop the pictureThe width or height of each frame.Whether important objects were removed at the edges.
Trim the sceneThe time interval included in the file.Whether the intended start and end remain.
Process an opening segmentOnly the confirmed beginning is included in this result.The credited range and actual output duration.

Verify the composition before adding new captions

Use the downloaded output as the source for the next edit.

Composition check

  • Compare the top, bottom and both side boundaries with the input.
  • Check a cup, face, hand or other subject close to the subtitle band.
  • Do not confuse black bars added by the player with pixels removed from the file.
  • Confirm that the exported duration includes the complete scene you intended.

Try your own video

Choose a local clip, preview it and review the cost before processing.

Free trial & upload limits

이메일 인증 후 체험용 크레딧 3개를 받으세요.

무료 계정: 50MB · 유료 계정: 100MB

1초 미만은 올림하여 초당 1크레딧 · 최소 3크레딧

자동 제거 시 다른 글자나 로고도 영향을 받을 수 있습니다. 요금 보기

Questions and answers

Will the result keep the same frame shape?

The workflow does not offer frame cropping. Check the exported file for the retained composition and for changes in the repaired area.

Is processing only the beginning the same as cropping the picture?

No. Prefix processing shortens the video in time; it does not cut a strip from the image.