Happy New Year to all Flame users around the world.
I’m currently working on a chroma key project and would like to get some insight from the community regarding keying workflows.
When doing chroma keying in Flame, it’s common to use tools like Additive Keyer to recover fine details such as hair. However, once these additional keys are layered on top—sometimes combined using different blend modes—they become difficult to extract or recreate later as a standalone key.
This brings me to my current issue.
For a new project, I requested the background and the chroma-shot subject to be delivered as separate elements, each with their own key. When I combine these two layers (background + subject with key), the resulting composite inevitably differs from my own master composite where everything was keyed and integrated together in a single setup.
My assumption is that this discrepancy is unavoidable due to how additional keys and blend modes are applied, but I wanted to ask if there is a proper workflow or best practice that allows the final result to match the original master composite more closely. Is there a recommended way to structure additive keys so they can be reproduced or transferred reliably?
Additionally, while there are many tutorials on forums and YouTube about creating clean plates, I’ve found relatively few resources that focus specifically on extracting a high-quality, accurate key. If anyone has recommendations for tutorials or talks that go deeper into advanced keying techniques in Flame, I would greatly appreciate it.
In particular, I’m very interested in techniques for pulling usable keys from chroma footage with heavy defocus or strong blur, where traditional keying methods tend to struggle.
Thanks in advance for any advice or resources you can share.
Every case is unique and will require you to learn whatever you can as there is no single bullet solution. Here are some awesome videos specifically for keying to get you started.
@mihran has a great setup that explains everything in detail
@PlaceYourBetts has been generous and included his setups, don’t forget to download them
@BrittCiampa does an excellent job of explain how and why image based keying works.
The hardest part of keying isn’t getting a good matte, it’s getting a good fill to push through that matte. Additive keys work great for comps, but you can’t export the fill and matte as a single RGBA image.
In order to achieve the RGBA “this can comp over anything and look fine” result, I extend the edges by painting them with the drag brush frame by frame. This is time consuming, but it’s gotten me out of many jams. I had a job once where the director shot the greenscreens on black and white film. We had a team of five painting motion blurred edges. The room was awash in the tap tap tap tap of wacom pens.
I’ve used pixel spread interpolate for this (shrink in matte, pixel spread out the edges first with stretch, then interpolate out from there.) Then a shrunken but generously soft edge to put the interior bits that aren’t going just be in that semi-transparent area of falloff back overtop. I’ll also this method to just fix spill contaminated edges that I don’t need to mess with the luminance of by using the interpolate layer set to photoshop color blend mode (WARNING: if working in Linear, switch to log for the PS color!!)
Thank you so much, this was a huge help.
I used to think that the legendary artists here could just apply a well-made preset and boom—pull a clean key even from blurred areas, but I guess it’s not that simple after all.
Thanks again, and wishing you a Happy New Year!
Thanks for the reply! @mihran
I watched the tutorial video a while ago, but I didn’t fully understand the last part. That might be because I’m Korean and watched it through translation.
In the final section, you push the background color quite aggressively and then use a key that’s not a clean “subject-only” matte, but rather a soft, hazy key that still includes some of the background, and then premultiply it back over the background—that part is what I’m referring to.
In real-world production, we often need to composite the subject back over the original background without changing the background’s original color. With this approach, it feels tricky because depending on the background, you then have to match and fine-tune that hazy key so it blends correctly with the original plate, which can become quite a headache.
It depends on your blend mode. If you have a layer that you’re adding on, create a second output from action, include that layer in the output of action, switch the output background to black, make sure the background actually has a value of 0(it may vary depending on colorspace), then you can export that output as a separate pass that can be brought into another program and added into a comp underneath your fill and core matte. Same process applies to multiply layers except with the background value should be 1. For overlay layers the background value should be .5.
I use the additive key method to just create the suppressed fill. The matte is the matte. The beauty of the additive key without using that as your comp is in the adaptive suppression process. So you’re not additive keying to a final comp. The way I go about it is to build the additive key, try to get my full looking close in terms of luminance and color to my BG that I’m actually going to comp over and then use that as just the fill. I think people get really stoked when they see how nice the additive key looks and think that’s the ticket, that’s the secret! BUT try doing some really refined color correction via client on your background (or as in your example the BG cannot change color) and good luck.
