taking one frame at a time since 1999

Hey guys,

A project came up where I'll have to do some green screen + video compositing in after effects/premier. This is not a subject I have experience with yet... Would anyone be able to share some resources on how I could learn this more in detail?

Thanks in advance!

Views: 96

Reply to This

Replies to This Discussion

I not very skilled with After Effects and only use it for a couple of things, but one of those is green screen keying.

There will be tutorials on Youtube, with far more details than i could provide, I'm sure.  But for the basics, I use a built-in plugin in AE called Keylight which works pretty well.  Here's how I do it, there are probably better ways, but this works for a non-expert like me:

1. Open After Effects.

2. Import image sequence.  Or movie file.

3. Drag it into the composition window.

4. Along the top of screen, click on Effect, and go down to Keying.  Slide over and go down list to Keylight (1.2).  Click on it.  Controls for Keylight will appear on the top left of screen.

5. Select the eyedropper tool and click on a typical part of the green screen.  The green will disappear - in my case, showing a black background colour, because I don't have any image layer underneath it.   

6. In the Keylight controls it will be showing View - Final Result.  Click on that and select Screen Matte from the drop-down list.  Now the image is black and white, with black being fully keyed out, and white being completely solid.  There will probably be some bits that are in-between, either greyish patches in the screen that are not fully keyed out, or darker bits in the white puppet area that are a bit transparent.

7.  On the list of controls, click on Screen Matte.  It opens up to show, under it, about 8 different controls.  The first one is Clip Black.  Default is zero, I slide it up and I see the black actually get blacker. Around 15 seems to make it solid black. (The white areas get darker too but don't worry about that yet.) At 17 the green cloth background is fully keyed out, but the green board the puppet is standing on has a lot of shadows, plus tiedown holes, and I have to go all the way up to 55 to get most of it to key.  I would need to manually paint out the white spots (not keying out) where the tiedown holes are.  (I normally give the puppet a bit of ground, coloured like the set, to stand on and cast shadows on, rather that make it green as well.)  I take Clip Black back to 17.  

8.  The next one is Clip White.  Default is 100.  I drag it down, and at 67 the white areas are completely white, so there is no see-through.  (That works with Clip Black at 17, but with Clip Black at 55 some dark areas start to re-appear in the puppet. That's because it is allowing too many shades of greenish to count as green and get keyed out.

9.  There are a couple of pale spots in the background, caused by dead pixels in my Canon DSLR.  I drag Screen Despot White up from 0 to, this time, 2.3.  Spots gone!

10.  Go back to View, and change back to Final Result.  You cn see how it looks in colour.  My background has keyed out.  I zoom in.  Then I change the View to Colour Correction Edges.  Yes, , if you zoom close there is a 2 pixel border around the puppet, it shows up as a grey outline, but it has been turned grey and does not show up at all.  If it does show up in the composite, there are Edge Colour Correction controls where you can grow the edges of the key by a pixel or 2 or 3, then do a softening so you don't get hard edges.  But usually I don't need to, the auto setting has done the job.

11.  Put the composition in the render queue and save.   I don't know if most people would layer the green screen shot over the background shot and save the composite in AE or not - what I do is save in a file format with Alpha channel so what I have is the puppet with a transparent background.  Then I can put that layer over another in another program (TV Paint Animation) that I am much more familiar with, and play around with size and position until I am happy. Maybe you can do that in Premiere, I don't know.  I only load finished HD clips into my editor (Da Vinci Resolve), with only a little colour grading left to do.

At his point the background is gone, but the green board with holes and shadows is not, I think the only way to get rid of it would be to paint out the whole area frame by frame - there is a black area around the puppet's feet so I wouldn't have to go right up against the edges of the puppet.

Before you get to post production work, set up and light the scene to give you the easiest key.  You probably know all this anyway.

1.  Avoid green colours on your foreground puppet or set pieces.  The further from green the easier it will be.  I've been doing some dino animation for Peter Montgomery's indie feature film (Dark Earth), and he sent me the resin dino head and body mould. I was to cast and make the body in foam latex, join the head on, and animate it. He painted it a very reddish brown, which I matched on the body, so it will be easier to key.  It is not the final colour - after the background is removed he shifts the colour back to a more greenish brown.  Coloured lighting on the puppet (but not the background) can also help to kill any greenish bits, and the colour balance can be corrected later.

Make the green background as evenly lit as possible, with soft front-on lights. Having the same colour of paint all over doesn't help if there are hot spots and shadows and fall-off, it will read as different shades of green.   It helps if you can put some distance between the puppet and the green background, to reduce green spill - light reflecting off the screen and back onto the puppet - and stop the puppet casting shadows on the green screen.  That creates more shades of green to key out.  However, if you have a green ground for the puppet to walk on, you can't separate them or avoid shadows, so avoiding greenish colours and shiny reflective surfaces on the puppet is even more important.

I do my keying in a large size, to get the best detail and edges, before reducing down to HD video size.  Images from a DSLR tend to have a greenish fringe around the puppets, where a border of 2 pixels will have a mix of the green background and the puppet colour. It's better to fix it in AE at the original high res, then scale the image down.

If anyone wants to chime in and say I am doing it wrong, or there is a better way, they are probably right.

You are a star Nick, thank you for writing this all down! I'll let you know how I get on.

StopmoNick said:

Nick has pretty much covered it all, but here's a tip I learned from someone who knows a lot more about VFX than me. If you have a subject which has an area that is quite close to the key colour and tends to start to go translucent when the key is about right, what you can do is duplicate the layer of your foreground subject, then make a mask around just the problematic area. Place this layer on top of the keyed layer and the interference issue is solved. Of course you may need to move the mask around, and feathering the edges is a good idea, but once you have the principle, it might make the keying a little easier as you can be more aggressive.

Sounds like a good trick Simon!  Now, if only I knew how to make a mask in AE...  or layers... i really only use AE as a last resort for those things I just can't do anywhere else.  I'm sure there are YouTube tutorials for masks in AE if I looked though.  

I'm sure the same thing can be done in Davinci Resolve's Fusion, which is what I am trying to learn just now. The layer stack is created in AE just by dragging an asset onto the timeline stack area. Then you can grab one and drag it up or down to reposition it in the stack. Top layer being the visible one. Duplicating is done by selecting a layer then pressing command+D. To make a mask, use the Pen tool in the top line and click to make a vertex.

I have heard DR's node system described as being like layers organised horizontally rather than vertically, with the added advantage that you can link things that are not necessarily directly stacked next to each other. AE has quite an awkward system for this, called 'parenting' - but then I suppose that's a fair comparison!

Thanks for the explanation Simon, which I don't understand a word of!  I find everything about AE is awkward (including the fact that using the forward and back arrows, as I do in every other program to move to the next frame, instead shifts the image sideways.)  I really should try to learn Fusion, since I am editing with Resolve anyway.   My access to AE is temporary - an old pre-subscription version on a 2008 computer that is getting too slow and is likely to fail soon, and we still have staff access to the current version from the university my partner used to work at.  One day they will tidy up and remove her account.  I can't justify paying Adobe's monthly fee for something I don't earn money from, so at that stage I would rather just ditch AE and find an alternative.  

Sorry if I am confusing. It was sort of an answer to your point 11. You grab a background plate from the assets imported into AE and place it as a layer underneath the one you have keyed. The whole composition, including both layers, can then be rendered into a single shot.

But as you say, it is probably more valuable to learn Fusion now. It took me about a year and a half of tearing my hair out before I got reasonably competent with AE, and I wouldn't want to put you through that!

Exactly the same process can be done in Fusion, with a background node as opposed to a layer attached upstream of the image node that gets keyed. The virtue being that when you go back to the edit page the composite image is shown as part of the edit timeline, without any rendering needing to be done until the final export. Here's a video

It can also be done in the Edit page by simply keying the image as you did, then placing it on the edit timeline above the background image. This is more like AE...

Hah!  I've been struggling with AE for about 15 years and have learned very little, and can't seem to retain it, so getting competent in a year and a half is better than I can do.   So it wasn't your explanation that was at fault, it was my lack of aptitude.  I really don't understand AE and find that if it can do something in a way that is counter to my intuition, it will.  So for things I can't do elsewhere, and need to do often, I have written down the instructions, and have to refer to them every time.  Fusion looks equally daunting and weird, but more worth the effort I think.  Can Fusion handle large 3x2 DSLR images like 5184 x 3856 pixels? Raw images?  Or do they need to be in a standard video format?  I convert image sequences to HD mov files before I put it in my editor, whether Final Cut Pro 6 or Resolve. 

One thing to remember to get a good key is to make sure that your project is set to work in floating-point. That means that your imagery is calculated at the highest possible bit depth. And in a non-destruktive workflow. It simply gives a lot more data to narrow down the green without eating up neighboring colors. You can always set the output to 24bit. But the calculations should be handled in 64bit or floating-point.

Happy keying :)

And to nick: If any program can handle large data it would be fusion. It will bring any computer to it's knees when set to render, but you can assign processors to the task in fusion, one of the rare benefits of having more. And while working on the project, you should set it to work with proxies, rather than full-size pictures. That's an in-program option, look it up. The full-frame you save for the final render. You should always import and export as single frames. that way if something goes wrong down the line, you haven't lost everything, but can just rerender the frames in question. Once you have all the frames .tif being the favorite, you can import those and make a Mov or Avi 

That being said - I've never tried to work with raw as input. If possible, it would give rather advanced possibilities in color grading.

Resolve can import frame stacks BTW pretty sure I've done so.

I'm having a zoom chat with Hans on Friday, as he clearly knows a great deal about all this.

I think you are right that Fusion is also daunting. Most of the problems are about where to find the wretched button that does X, and the only way to learn both that and AE seems to be to blunder around until it starts to yield. 

Davinci Resolve has one advantage over the Adobe programs, as far as I can see. This is that you bring in your images (big as you like if your computer can handle things) at the beginning, then do all the editing, VFX, sound and colour correcting before ever exporting anything. So you just have the single export to whatever you are intending to show it on. This would preserve maximum quality. BUT, as I understand it, unless one has a very up to date and powerful computer, the program will soon start to stutter if you use the big files right the way through.

So one sets up proxy files, which are tiny, do all the work on them and then at the end the program swaps out the proxies for the big files and renders them to video output.

That is my slightly simplistic understanding. Of course when doing something like removing rigs you need to work on the big files, but then one does not need smooth playback. I hope Hans will help to clarify for me about all this, and things like floating point, which I don't really grasp at all. In fact things like color space are a bit of a mystery too.

One tip: if you are importing jpegs into DR, there is a little drop-down three line button at the top L of the media input section which allows you to choose to import as single frames or as a sequence. I got a bit stuck with storyboard stills all coming in as a single clip instead of individually.

StopmoNick said:

Hah!  I've been struggling with AE for about 15 years and have learned very little, and can't seem to retain it, so getting competent in a year and a half is better than I can do.   So it wasn't your explanation that was at fault, it was my lack of aptitude.  I really don't understand AE and find that if it can do something in a way that is counter to my intuition, it will.  So for things I can't do elsewhere, and need to do often, I have written down the instructions, and have to refer to them every time.  Fusion looks equally daunting and weird, but more worth the effort I think.  Can Fusion handle large 3x2 DSLR images like 5184 x 3856 pixels? Raw images?  Or do they need to be in a standard video format?  I convert image sequences to HD mov files before I put it in my editor, whether Final Cut Pro 6 or Resolve. 

I need to do different things to different shots, and want to put the HD shot into the edit as I finish each shot, so I doubt I will be importing them once as large jpegs and doing everything there in Fusion.  For some, I do the rig removal in TV Paint, then keep them as large 3x2 images so I can zoom in and pan around using the virtual camera in Lightwave. It has good ease-in and ease-out controls on camera moves. With other shots, I  get them down to 1920 x 1080.  Wouldn't I have to create different projects in Fusion to handle the different sizes, then export, so my actual final edit project only imports 1080 HD?

I am doing very similar things, and it bugs me to have to import and export shots between programs, as this is where quality gets lost.

As far as I know Davinci Resolve addresses this issue by providing a complete suite of tools within a single program. Fusion is only one part of this, so if you import a stack of JPEGs into DR through the Media page (on the left at the bottom of the screen) you can then do a rough edit in either the Cut page or the Edit page (which looks most like Premiere Pro). At any point you can highlight a clip and go to the Fusion page (the equivalent of exporting from PPro into AE). In Fusion you can then do rig removal, VFX etc. When you go back to the Edit page the adjustments to the clip are visible in the timeline - without having to do any rendering at all. Colour correcting can also be done in this way simply by clicking on the Color page. There the highlighted clip shows up in a timeline so one can match it to surrounding clips. Same deal with the Audio page.

It does seem to be a one-stop shop....and I am just describing a JPEG sequence imported and not converted to anything. The project dictates the size, so one can select HD or even 4K. I understand it may even be possible to open a project in HD but then decide later to export in 4K. The point being that the imported footage is only sized to the project in the Deliver page, not at the beginning.

Of course with this scenario there are snags. Unless one has a fast modern computer playback will be stuttery in HD and jam up completely in 4K, because what the machine does is look at all the instructions for each frame before displaying it and moving in to the next one. With essentially four programs in one the workload is much higher. So this is where optimized media and/or proxies (same thing, one is just automatic) come in. They enable you to edit using a relatively small file size, with of course a loss of detail in the display. When the time comes to export from DR all the proxies are automatically swapped out for the hi Res files and the adjustments applied.

I think it is possible to swap back to the original hi Res files for e.g. rig removal, so one can get the necessary detail. And Hans is talking about floating point colour, which my limited understanding tells me is about having the maximum colours available per pixel.

Cautionary Note: what I have described above is what I have come to understand is offered by DR. I am still very much at the shallow end of using it, and have not done much more than make a few YouTube videos with it. Some of the processes are quite different from the Adobe style, so need a lot more learning. For example I am trying to do some roughly animated storyboard sketches. I start with the drawing in Photoshop, cut out the figures I want to move and make separate layers for background and figures etc. Then I do a quick export to PNG and import those to DR. I tried to do my moving around in Fusion (I used AE before), but it works much better in the Edit page, where I make a stack on the video layers and then highlight each PNG. The Inspector tab comes up (top R) and in there I can move and keyframe the images around. Once I have done all that to my heart's content I highlight the whole stack and right click, which brings up a tab. Near the top is Create Compound Clip. Clicking on this reduces it all down to a single line. I can adjust the timings at any time by opening the compound clip and fiddling around. It is the direct equivalent of pre-comping or nested clips in AE (but I guess that might not mean much unless you use it...)

At the moment my animatic project in DR has all my scenes in it, and each one is a separate compound clip, so I can open them individually and mess around with them, then close them and they appear in the general timeline. One hiccup is that DR does not show the extra bit of clip if I have lengthened it, and if the scene's compound clip was butted up against the following scene on the same video track in the timeline it would get chopped short. So I overcome this by alternating the clips on video tracks 1 and 2.

That's probably enough for now! I have high hopes that DR will make things a lot easier in the long run, but it does take some getting used to.

Reply to Discussion


Website News & Tips

This time, the theme is HALLOWEEN. All entries must be submitted by September 18, 2021. More info here.

If you are new to the site, please feel free to introduce yourself on 
this Message Board thread.

ou can now customize your personal page. Go to My Page/Customize My Page.

To control who sees your posts and Profile Page, go to Settings/Privacy.

If you have a question for the community, it is best to post your question on the Message Board instead of a Blog. Blogs are best for featuring your stop motion projects.

When posting a video, if possible, use the Embed Code instead of the URL. Your video will be a larger, more viewable size.


If you have a question or are looking for answers, you'll find a wealth of information here on the Message Board.

BEGINNERS a good place to start
ANIMATOR TALK experienced animators
YOUR STOPMO FILM PROJECT what r u working on?
STORY script, storyboarding and storyreel
 what's a surface gage?
CAMERA & STAGE camera, lighting and moco rigs
SOUND lip-sync, sound effects and music 
ARMATURES ball & socket and wire 
MACHINE SHOP metalwork tool & talk 
SCULPTING sculpture information and advice 
HAIR & COSTUME materials and technique 
 foam, silicone and resin 
CLAY clay puppet construction 
other puppet fab issues
STOP MOTION SETS set design and construction
 miniature prop discussion

and even more...



© 2021   Created by Anthony Scott.   Powered by

Badges  |  Report an Issue  |  Terms of Service

.xg_widget_forum_index_index .xg_column.xg_span-7.xg_last {display:none;} .xg_widget_forum_topic_listForCategory .xg_column.xg_span-7.xg_last {display:none;} .xg_widget_forum_topic_show .xg_column.xg_span-7.xg_last {display:none;}