-
Proper workflow advice for difficult roto/track/key/matte/mask/
EDIT— i didnt mean to suggest that i havent yet tracked. I have. in mocha and AE… many dozens of times… i just dont know how to I guess have it end up maybe as a procedural matte???? or somethin
Hello, folks. So, here is the deal, veals… Over the last five years or so I have pretty much never posted a question to ya’ll but i have gotten answers from you, many many dozens ot times, as this is always my first click after a google search. So thank you for the past help. I am an aspie and so this is out of my comfort zone.
I am not a professional compositer,, vfxr, editer, etc-terer, but I am far from a moob. But I am having roto issues and I am sure that tracking can solve it, I am just also sure that I have been going about it all the wrong way, due to missing maybe only one piece of info. Maybe more…
My task is roto-ing and or keying out fast moving foreground dancers from their backgrounds, which are generally shot with a tripod… Difference matting usually yields venus flytrap type resutls and only gets me so far. I have done color grading to enhance contrast at times, but I am always, always left with no more contrast left between the dancer/actor and multiple background elements.
Auto trace is a joke, alone, and as tricky as diff mattes… So, I know tracking is the way to go, especially cause there are usually 8000+ frames, every flippin video, and manual roto is driving me crazy. The roto brush is just not going to happen, that damn propegation somehow always flippinj disappears at some ‘random” thousandth frame, and so super lame that ctl+z wont bring it back. Wasted a dozen hours maybe oin that..
I cannot find ANY relatively relevant tuts disregarding the “track the guy running in front of wall” default tut, and that isnt doing it at all for me. ..
While I initially thought thatt tracking the actor was the best way to go, I have just started thinking that a holdout matte/mask in Mocha for background elements and then maybe the auto traced actor with the mask set to intersect would allow just “him” to be shown when moving in front of objects…… is the way to go…. maybe….. ????
I mean I will do whatever way is “best”, period. I just know that I dont have all the information needed for me to understand the best way to do this…Clearly
Maybe just the workflow that has been proven to result in the desired results would be silly sicc and so appreciated…. Im DONE with pushing out chewy ch!t that can result in the opposite result and expereience for the viewer than I am looking for. I dont even need it “perfect”, as the vfx and atmosphere will hide edge imperfections. But holes and jagged crunchy chewy crustyness are killin me, loudly…stabilization too is a must yes?
Maybe ta Rowland willl come rollin through this thread and say bam bam bam and link me to the bam and I can relax again so that I can work more and more…. Here is a demo reel showing my past chewyness.. But as it is a new compound art form, I was less concerned abiout the appearance than the abilities being conveyed and wasnt gonna let anyone else touch them with hands on… But I have like 8 or ten more to do, myself, and this is the key to it all, i think.. Below is demo and examples of the types of fn bs garbage I keep getting tstuck with… Thanks, sorry for length….
Infinity Edition Intros 1 and 2 and crossover breakdown
Silver Slapper
Some contents or functionalities here are not available due to your cookie preferences!This happens because the functionality/content marked as “Google Youtube” uses cookies that you choosed to keep disabled. In order to view this content or use this functionality, please enable cookies: click here to open your cookie preferences.
Starcrossed Slappers (iTernity and Mistress Death)
Some contents or functionalities here are not available due to your cookie preferences!This happens because the functionality/content marked as “Google Youtube” uses cookies that you choosed to keep disabled. In order to view this content or use this functionality, please enable cookies: click here to open your cookie preferences.
Crossover and storyline breakdown
Some contents or functionalities here are not available due to your cookie preferences!This happens because the functionality/content marked as “Google Youtube” uses cookies that you choosed to keep disabled. In order to view this content or use this functionality, please enable cookies: click here to open your cookie preferences.
Blood type O negative. I bleed ones and zeros
I dabble on the desktop
Shifted Leering, Any Perspective
Shifting Literally All Perceptions
SLAP -
10 Replies
-
Mocha is probably the way to go with this one. I know you mentioned that you have done it before, but what is your basic workflow? Are you only using one spline mask?
My suggestion would be to break the body up into 12 spline masks, and track them each on their own. and copying the mask information back your footage. Something like: Hands, forearms, upper arms, head, torso, upper legs, lower legs, feet. That way you can be more definite on your masks and not try to have the program track too much at a time.
Just a thought.
Best,
Derek Boettcher
Creative Management
BTH Media Group
https://bthmedia.com -
Thanks, Derek, appreciate it. My workflow has generally been to do a track of the torso across the footage. Two, usually. Then I always seem to go for a thigh.
I then try to create a linked layer as a mask for the thigh. But most of the time it will not act parented and not sure why, as I thought it was only selecting the linked layer in the drop down menu. That must be user error, though. Perhaps I altered a setting, I will default it.
I hear what you are saying about the twelve tracks but there is still the HUGE issue of the background not contrasting enough to be extracted/keyed. Are you suggesting that it is a better idea to track what I want instead of tracking what I dont want, only? And also, I am usually tracking after I have removed 90% of the background. Is this more or less effective than a track with the full footage?
I think, also, that maybe a script that would allow me to associate a track point with a mask vertex may do big things, too, as the auto trace will get “most” of the edges. If I have a auto trace of the overall shape and path of the layer, is a mocha track even necessary?
IS there a script or ability to link a track point or even a masked shape layer to a set of specific vertexes in an auto traced mask?
You asked for the workflow. It is after that third track or whatever that I usually just go wtf am I doing and move to another aspect of it. Clearly have some stupid disconnect happening..
For each actor i generally have a half dozen rendered out footage sequences. All differ do to my attempts to remove the background. But I cannot seem to visualize how to break those things apart in order to just have the stuff necessary fit and the stuff that isnt, removed.
And on the twelve track aspect, when I get to the “ninth” track, will I be using the previous eight as reference, for speed and such or are all independent with each other? Cause that is over 100 thousand frames, per video, with twelve tracks..
Blood type O negative. I bleed ones and zeros
I dabble on the desktop
Shifted Leering, Any Perspective
Shifting Literally All Perceptions
SLAP -
Hello, again. Okay, so here is actually a perfect example of something that is giving me issues. And I think it is because there are so many ways to get it done, I just dont know which is most effective..
I need to have his head and hair filled in, obviously… edit—grr, okay rigid mask tracker worked well for this one, but wont work most times, this had good contrast between clothes and background and nothing to interfere with track…
But the 12— I KNOW that there has to be an answer with auto trace, trackerviz (maybe masktracker+, too) and either mocha or AE tracker. It seems to me that even if i got an auto trace of this actor, with feet, hands, and head removed with the background, there is enough data, more than enough, with the mask vertexes, to not have to do much more in order to parent something to them somehow and just go off that..right? Cause each vertex on the mask is essentially a track point, and “100”
track points is WAY way enough to be able to do much more complex linking of properties and objects…Does anyone have in depth understanding of trackerviz? The tuts only deal with corner pinning, like maybe 10% of the options/menu items are covered and the dood isn’t replying to my post on aeenhancers…. I think this script is probably the most powerful script that “no one’ either fully grasps, keeps it their super special secret, or has no reason to use it for things other than corner pinning..
Grr, so frustrating when I have so many choices and no foundation set that would enable me to pick the right path, initially…
Blood type O negative. I bleed ones and zeros
I dabble on the desktop
Shifted Leering, Any Perspective
Shifting Literally All Perceptions
SLAP -
Better example… Let us call the blue on the edges the auto trace. Shouldnt I be able to , am I able to , create the mask that would cover the head and then lock all the vertexes where blue and red meet, , where the green arrows are pointing?
I cannot figure out how to do this, effectively. I keep having to use mocha or built in tracker at this point, but i thik that is just extra steps that arent necessary with all the mask points…
Blood type O negative. I bleed ones and zeros
I dabble on the desktop
Shifted Leering, Any Perspective
Shifting Literally All Perceptions
SLAP -
I dont know what happened hiere– earlier post that didnt get posted, I guess…
Hello, again. Okay, so here is actually a perfect example of something that is giving me issues. And I think it is because there are so many ways to get it done, I just dont know which is most effective..
I need to have his head and hair filled in, obviously… edit—grr, okay rigid mask tracker worked well for this one, but wont work most times, this had good contrast between clothes and background and nothing to interfere with track…
But the 12— I KNOW that there has to be an answer with auto trace, trackerviz (maybe masktracker+, too) and either mocha or AE tracker. It seems to me that even if i got an auto trace of this actor, with feet, hands, and head removed with the background, there is enough data, more than enough, with the mask vertexes, to not have to do much more in order to parent something to them and just go off that..i believe.. Trackerviz and Cause each vertex on the mask is essentially a track point, and “100” track points is WAY way enough to be able to do much more complex linking of properties and objects…
Any with in depth understanding of trackerviz? The tuts only deal with corner pinning, like maybe 10% of the options/menu items are covered and the dood isn’t replying to my post on aeenhancers…. I think this script is probably the most powerful script that “no one’ either fully grasps, keeps it their super special secret, or has no reason to use it for things other than corner pinning..
Grr, so frustrating when I have so many choices and no foundation set that would enable me to pick the right path, initially…l
I would think that 12 tracks of the footage may take me twelve hours. Man what I would have given for only twelve hours instead of 40+ trying to get the perfect key without greens creen
———————————————–
Blood type O negative. I bleed zeros and onesI dabble on the desktop
Shifted Leering, Any Perspective
Shifting Literally All Perceptions
SLAP -
Well the main problem I see is that when you try to key, you end up keying out colors from the background that are naturally on the talent, (given why green screen and blue screens are so effective). In my opinion, mocha’s planar tracker is the way to go with this; it can just be a pain because you have to create multiple track mattes for one body. But the majority of the work would be fairly automated and you should only be to do minimal key frames. In the future of you have control over it,I highly suggest shooting on solid wall (preferably green or at least whatever is not on the talent) you will then get a much stronger key. I’ll look around to see if I can find settings that might help. Can you upload a sample of the video? I’ll see what I can come up with
Best,
Derek Boettcher
Creative Management
BTH Media Group
https://bthmedia.com -
Hi, yes, I will upload. Thank you. And I fully agree about the green/blue screen. All of these footages were shot over a year ago so I had no choice in the matter.
And I understand your point about the keying, but I am struggling to grasp why, for example, if I am able to get an auto trace mask that more or less remains on body shape paths the entire length of the footage, in the areas where there are holes or caverns down from edges, why i cannot select the closest “5” mask vertexes and attach a shape or a layer mask to those vertexes and have it base its movements on them… group parenting, or whatever.
With almost anything, really, if you have 90% of the information, most times you can fill in the rest of the information based on relative data.
I, also, may be using “difficult” in my own way, and it may be that these things are not difficult for folks such as yourself and wouldnt require 12 tracks. Ill upload two samples tonight and see what you suggest. Thanks again, appreciate the help.
———————————————–
Blood type O negative. I bleed zeros and onesI dabble on the desktop
Shifted Leering, Any Perspective
Shifting Literally All Perceptions
SLAP -
Hi, yes, I will upload. Thank you. And I fully agree about the green/blue screen. All of these footages were shot over a year ago so I had no choice in the matter.
And I understand your point about the keying, but I am struggling to grasp why, for example, if I am able to get an auto trace mask that more or less remains on body shape paths the entire length of the footage, in the areas where there are holes or caverns down from edges, why i cannot select the closest “5” mask vertexes and attach a shape or a layer mask to those vertexes and have it base its movements on them… group parenting, or whatever.
With almost anything, really, if you have 90% of the information, most times you can fill in the rest of the information based on relative data.
EDIT… I think I have also been making a huge mistake by not stabilizing the footage no the actor.. I am doing a few tracks in Mocha, right now, and will see if that makes a difference.
———————————————–
Blood type O negative. I bleed zeros and onesI dabble on the desktop
Shifted Leering, Any Perspective
Shifting Literally All Perceptions
SLAP -
Hello. Here is the original footage
Some contents or functionalities here are not available due to your cookie preferences!This happens because the functionality/content marked as “Google Youtube” uses cookies that you choosed to keep disabled. In order to view this content or use this functionality, please enable cookies: click here to open your cookie preferences.
I just uploaded a couple of my partial keys, too, but realized just now that they ar mp4 with no alpha. Will have to upload an image sequence right now
black w/o alpha
https://www.dropbox.com/s/8nyou71f6dswvua/Pre-comp%205.zip
alternate key
https://www.dropbox.com/s/8dkzsxjrvxngcfy/raino_legmatte_1.zipThanks again, fun times and all…
———————————————–
Blood type O negative. I bleed zeros and onesI dabble on the desktop
Shifted Leering, Any Perspective
Shifting Literally All Perceptions
SLAP -
Hello. So, mocha is giving me problems with parented layers going crazy with scaling. i created another thread here https://forums.creativecow.net/readpost/2/1059920
This is a problem I have had anumber of times with Mocha and no idea why it happens
———————————————–
Blood type O negative. I bleed zeros and onesI dabble on the desktop
Shifted Leering, Any Perspective
Shifting Literally All Perceptions
SLAP
Reply to this Discussion! Login or Sign Up