Andrew Somers
Forum Replies Created
-
I may have skipped over a bit above:
When using the UNMULT plugin, in combination with the gamma/blur technique, You would place the UNMULT plugin at the top of the stack, then CHANNEL BLUR and blur only the alpha channel, and then LEVELS set to ALPHA, and adjust the Alpha’s GAMMA.
Use the amount of blur and the alpha’s gamma to adjust the edge (i.e. choke/softedge etc). Adjusting the alpha’s input black and white point will additionally adjust the density/opacity.
This would generally be for a layer using NORMAL transfer mode.
-
To add to the link Kevin listed to my earlier posts, I just want to mention that there is a free plugin available at Red Giant that can be helpful and can eliminate some of the steps I indicated in that earlier post especially for transparent items like smoke.
It’s called UNMULT
https://www.redgiantsoftware.com/products/all/knoll-unmult-free/
It’s free and I highly recommend keeping it in your arsenal of tools.
I’ll use Unmult to pull an alpha by unmultiplying an object against black (though I will still use the gamma/blur techniques I discussed in the link Kevin posted to adjust the alpha edge). But Unmult really needs a *solid* black – if your black is not at 0, then you may still need to do some of the tricks indicated on that link.
Black screen is best for:
Objects that are transparent such as smoke or water.
Objects that are well lit and have very little shadow in them.For instance, a blonde with backlit hair may be a good candidate for black screen, but on the other hand, something like a wine glass may not – while we think of a wine glass as being “transparent”, the caustics of the glass actually cause quite a lot of internal shadow and this may make them a bit difficult against black screen.
Things that are *dark* and should be *opaque* are bad candidates for black screen (though in some cases, that kind of object might be good in white screen).
Getting to the OP’s question: Blood Splats.
Blood is so dark and you usually want it fairly opaque, I don’t think I like the idea of shooting it against black screen.
If you are using something from a library, there *should* be a version with an alpha built-in.
If you are shooting this yourself, and you want to use blackscreen techniques, then don’t shoot “dark red blood” INSTEAD, try shooting milk or cream, or flour mixed with a lot of water, then use the TINT plugin to tint it in AE to the desired color. Light it so that you have the appropriate specular highlights and shadow modeling. Keep the *shadow* at least two stops brighter than your black screen, and get the black of the black screen as close as possible to the clipping point.
ALSO: The techniques I discuss here and in the previous thread essentially require that you be working in 32 bit LINEAR colorspaces. IMO working on a gamma encoded space gets in the way.
-
Just as an FYI. BouJou is very expensive ($10,000), and slow, and may be overkill in many situations – but if you have a shot that no other app can solve, BouJou is the one that can.
I have yet to try the new 3d camera solver in CS6 so I can’t compare, BUT:
If you are new to 3D solving, consider:
PFHoe – a cheap $100 app that is a very stripped down version of PFTrack ($3300). PFHoe has a number of limitations, but is also a great way to get your feet wet in 3D Camera Solving. For simple shots, it does a surprisingly good job.
The Foundry also has a 3D tracking plugin for AE that is reasonably priced.
And the standalone SynthEyes is also a very capable plug in at around $600.
Also, the new Mocha 3 now adds 3D camera solving to their excellent planar tracking methodology, and I believe it is included in the very reasonably priced Mocha3 AE.
It is important to note that each 3D camera solving application has their own idiosyncrasies and workflows, and one may outperform another in a certain situation. I have not upgraded to CS6 yet but am interested to see how useful their 3D camera solver is.
-
sanex said: “if you have all the footage done it doesnt take a lot of time”
LOL yea right.
In the first minute of the above video, there are 16 separate camera tracking shots or object tracks. And at least half a dozen animated rotos. It would take more than two working days just to complete THAT work for just the first minute of the video, and that’s not including any of the graphics and compositing or color correction.
The term “working days” is also kind of nebulous. Do you mean a normal non-union 8 hour day? Or are you working 16-18 hours to try and shove this out in “two days”???
THE REALITY:It is best to consider the total “hours” that such a task will take. To do this break it down into the various segments/ component shots and working tasks.
Assuming that you already have an offline cut as your starting point then a task list might look something like this (workflow assumes AE CS6 and the bundled version of Mocha:
TASKS:•Pull counts for footage from offline.
•Load and Lineup camera originals into After Effects sequence(s).
•Determine the type of track needed for each shot. For instance, some of the shots of the text next floating next to the person might work as a planar track in mocha. Others though clearly need a 3D camera solve.
• Determine the other tasks per shot such as tracked roto, kind of graphics animation, etc. For instance, a person walking in front of a graphic will need a tracked/animated roto created in Mocha. Some of the graphics in the above video needed some level of 3D work, while others worked as a simpler 2D graphic.Once you have the breakdown, then you can estimate the time involved.
CONSIDERATIONS:
3D tracks:
If you are creating a 3D camera, remember that once you starting building the shot with the camera, than any adjustments to the camera solve will require significant changes to the 3D environment/elements to match the new solve. As a result it is very important that you assess that the solve is stable before doing significant work.
As such, after completing a solve in PFtrack of BouJou, import the camera and then use the nulls as validation points, and render to see if it is as stable as you need. WHile there are tools to validate inside PF or BJ, it’s good to double check inside AE. If using the new 3D solver in AE6, it will be important to validate the quality of the track before doing significant work.
Therefore, it is important to time budget for the time needed to validate a 3D solve and adjust. It’s bad to assume that because an autotrack and autosolve takes a total of 30 minutes, that you only need to allocate 30 minutes to the task. In the real world, an autotrack and autosolve will rarely work on the first pass, and will need substantial massaging to get it to work.
A *simple* 3D solve should be expected to take no less than 2 hours (including validation). Some complicated shots can take days to get a solid, useable solve.
If doing 3D tracks, you want to have a second system available to continue other working during solve/adjustment iterations. (Or enough RAM to have multiple large applications running simul).
Animations:
My generally feeling on planning the time budget for animation is to follow this formula:
TimeToComplete = (AnimationElements * ShotLength * how_long_you_think_it_should_take * yourDesireToSleep) * 10
I.e. – animation can take up a vast amount of time, not just in the setup and conceptualization, but in the little tweaks and finesses – you render, see something you want to massage, go in fix it and re-render – but wait there;s a nothing little thing….It is a very iterative process. What was done in the above paris video looks simple and elegant – but that apparent simplicity undoubtedly took many iterations to finesse.
Roto:
Mocha is a great tool for creating animated, tracked rotos – but even so, it still takes quite a bit of time and labor to get good solid and usable rotos.
SUMMARY:
At any rate Estelle, I hope this was useful – If I were to pull a figure out of thin air without doing a proper breakdown, I’d say that the above Paris video was somewhere in the 200-hours of work range.
-
And you would be failing to describe what exactly it is you are trying to do, now wouldn’t you?
-
On review of your sample:
You DO have banding in the original, and it is just that RSMB is “enhancing” it. Note that even applying a simple gaussian blur will bring out the banding. THE REASON: The original has dithering that hides the banding, but adding a blur tends to smooth out the dithering, and this makes the banding easier to see.
The banding in the original is very hard to see but is there – though to be certain, I’d rather look at some of your original PNG files.
TO FIX try one of these solutions:
A) Add noise AFTER the RSMB plugin. a tiny amount of noise will result in dithering and reduce the appearance of banding. Add the NOISE plugin, and try setting it to a very small amount like 0.2%, color OFF and clip OFF.
OR
B) Re-render in C4D to EXR files and use LINEAR light mode in C4D, then work in AE in 32 bit LINEAR mode. When you bring in the EXR files, they will also be in linear mode.
-
You need a hardware/software combination.
The HARDWARE I recommend is the X-Rite display.
I had the XRtie Eye One Display 2 (still available), but I just bought the Xrite EODIS3 (waiting for it to arrive later this week)
My original i1 Display 2 was pretty decent, though the included software was fairly weak.
The *best* software for calibration is BasICColor Display, and it works well with the XRite calibration devices.
https://www.basiccolor.de/basiccolor-display-5-en/
The *new* XRite that I ust bought comes with new XRite software that is supposed to be improved, though I have not tested it yet.
FWIW, I find that the “ColorEyes Display” software is bad and inaccurate (and their support is terrible), and also the Datacolor Spyder devices do not match the performance of the XRite products.
My suggestion then would be to buy the new XRite:
https://www.amazon.com/Xrite-EODIS3-i1Display-Pro/dp/B0055MBQOW/ref=pd_cp_pc_0
And possibly add the BasICColor software later if you really want to “tweak”.
-
If you are talking about stereo 3D – well there are more than a few but they are very expensive, high end solutions.
If you are talking about taking a 2D image and doing the “Ken Burns” effect of making it 2.5D, that is a different process/concept all together. In that case you might want to check out Digieffects Depth Cue
https://www.digieffects.com/product/buena_depth_cue#tutorials
-
Thanks ADam that is useful – a followup question: I am finding that the apparent root of he problem is a number of cases where multiple polys are sharing the same vertex, including cases where the outside polys and the inside polys got too close and a common vertex merged.
Is there a way to split a vertex, to separate several polys without deleting the polys themselves?
-
The After Effects tracker probably won’t do what you need – it sounds like you will need a true 3D track, and then import the 3D camera into After Effects.
Though you *may* be able fake what you want – using expressions with the “toComp” function may help:
https://help.adobe.com/en_US/AfterEffects/9.0/WS3878526689cb91655866c1103906c6dea-79fda.html
https://provideocoalition.com/index.php/cmg_keyframes/story/dmoe4_space_conversions/
And also, a MOCHA track might get you close enough that you may be able to offset some planar tracks to get the look you are going for.