• Global community
    • Language:
      • Deutsch
      • English
      • Español
      • Français
      • Português
  • 日本語コミュニティ
    Dedicated community for Japanese speakers
  • 한국 커뮤니티
    Dedicated community for Korean speakers
Exit
2

Remove one person from a video clip, from a background full of dancing people

New Here ,
Jun 03, 2023 Jun 03, 2023

Copy link to clipboard

Copied

Is it possible to use content-aware fill in after effects, to remove one person in the foreground of a video clip, with lots of people dancing behind them?  Ive read that this feature only works with very unbusy basic backgrounds.  The video was filmed in slow-mo.  Has anyone ever managed it and how? 

TOPICS
How to

Views

947

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Community Expert ,
Jun 03, 2023 Jun 03, 2023

Copy link to clipboard

Copied

It's impossible to make an educated comment without seeing the footage, or at least a still from the footage.  But what you've read is generally true - Content Aware Fill is quite good at replacing static and slow moving content, but can struggle when there is a lot of movement/change.

 

It's quite a straightforward process, so your best bet is to just give it a try.

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
New Here ,
Jun 03, 2023 Jun 03, 2023

Copy link to clipboard

Copied

Screenshot 2023-06-04 at 03.53.24.png

here's a screen shot, cant seem to upload clip ... The guy in the centre, needs to removed bc he gatecrashed the event, was a pretty wasted and was wearing a labelled top, but its a great slow mo shot, so wanted to use if at all possible. Thanks for advice, Im managing to make mask, track it, but it seems the background gets messed up.  Also, I cant seem to export the final results. 

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Community Expert ,
Jun 03, 2023 Jun 03, 2023

Copy link to clipboard

Copied

Unfortunately the scene looks too complex for CAF to be effective.  I did a CAF on your single frame and it's obviously a bit of a mess.

Screenshot 2023-06-04 at 1.25.45 pm.png

 

 

But you could potentially save the shot by getting creative.  If you can use it in your edit as a background for a title, or something, you can draw the viewer's eye away from the problem area.  Just a thought, without knowing what your footage is really like.

 

Screenshot 2023-06-04 at 1.26.02 pm.png

 

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Community Expert ,
Jun 03, 2023 Jun 03, 2023

Copy link to clipboard

Copied

Had the person obscured something that's easy to replace like a wall or bushes, Content-Aware Fill could have been goood for removing the person.  

 

If the "SIVENCHI" on the shirt is the primary issue, Content-Aware Fill might do a good job removing that.  If not happy with those results, good third-party options include Lockdown and Mocha Pro Power Mesh.  

 

If the person does not move much, Detail-preserving Upscale could be used to double the frame size of the clip and allow the shot to be reframed to show the crowd to the left.


Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Community Expert ,
Jun 04, 2023 Jun 04, 2023

Copy link to clipboard

Copied

LATEST

Content-Aware Fill requires one object that moves across a background that does not change. A drone shot of a car driving down a highway works because there are frames before and after the car passes mile post 25 that do not have the car at mile post 25. 

 

If you have a moving camera, a moving crowd of people, and you want to remove one person, you need some frames before and after the person moves in the frame that have the other people in in the same position they were in before or after the person that was moving got there. That's asking a lot from any kind artificial intelligence. The likelihood of it working automatically is astronomical. 

 

 

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines