Hi,
I just downloaded the MediaStudio Pro Trial. I am familiar with Adobe Premiere and After Effects, and I am hoping that this software will work as a good replacement for both. Unfortunately, I don't have a lot of time to check out the trial.
I do 3d animation, and I would like to composite separately rendered "animation layers" with pre-multiplied alpha channels (specular passes, shadow passes, reflection passes, etc). I have not been able to figure out how to do this in MediaStudio Pro. Here are my questions:
1. I can't find a way to import image sequences directly. Do I have to convert every image sequence to a .uis file and then import those as videos?
2. I can't seem to get my .tif alphas to show up in MediaStudio Pro. I used the above method of first creating a .uis. Did that get rid of the alphas?
3. Is MediaStudio Pro a good option for this type of compositing? How does it compare to After Effects in terms of cababilities?
Thanks in advance,
Shawn
Compositing Capabilities?
-
NickJushchyshyn
I use MSP for this type of compositing quite often. It works very well EXCEPT, MSP is not really setup to handle pre-multiplied alpha very well. A premulitplied alpha render will typically leave a line around the edge of your objects. Most CGI software will output non-premultiplied renders and I'd recommend this form of output regarless of your compositing tool. I know that more advanced compositing applications like Shake (costs a couple thousand dollars) have a setting that will compensate for pre-multiplied renders, but this is not a feature of MSP (<$300 
)
MSP is NOT as powerful for compositing as AE. AE is a tool DEDICATED to compositing, but doesn't handle video editing very well. (I also use AE all the time. If I'm working on a complex comp, AE gets the task. If it's just a basic comp of a multipass render to be included in a broader video, I can through it at MSP)
A UIS file is not a "converstion". It's only a description file that "explains" the image sequence to the video editor in MSP. It takes only a couple of seconds to write this file and you can create this file interactively directly in the "Insert Video" dialog. This is not a whole lot different than what AE does internally in it's project file when you bring an image sequence into your project.
Once placed on the timeline, your alpha data is STILL preserved in your TIF files. In fact MSP is reading this TIF files natively and has not altered them in any way. To make use of the alpha channel, simply right-click the TIF clip in your timeline, select Overlay Options, and use the resulting dialog box to set the Overlay Type to ALPHA.
For an ambiant pass you can use a mulitply overlay type over your TIF track. Specular passes work well as an "add" overly type. You can use the gama and cutoff controls in the "advanced controls" area of the overlay dialog to "tune" these passes to your liking.
Hope this helps.
)MSP is NOT as powerful for compositing as AE. AE is a tool DEDICATED to compositing, but doesn't handle video editing very well. (I also use AE all the time. If I'm working on a complex comp, AE gets the task. If it's just a basic comp of a multipass render to be included in a broader video, I can through it at MSP)
A UIS file is not a "converstion". It's only a description file that "explains" the image sequence to the video editor in MSP. It takes only a couple of seconds to write this file and you can create this file interactively directly in the "Insert Video" dialog. This is not a whole lot different than what AE does internally in it's project file when you bring an image sequence into your project.
Once placed on the timeline, your alpha data is STILL preserved in your TIF files. In fact MSP is reading this TIF files natively and has not altered them in any way. To make use of the alpha channel, simply right-click the TIF clip in your timeline, select Overlay Options, and use the resulting dialog box to set the Overlay Type to ALPHA.
For an ambiant pass you can use a mulitply overlay type over your TIF track. Specular passes work well as an "add" overly type. You can use the gama and cutoff controls in the "advanced controls" area of the overlay dialog to "tune" these passes to your liking.
Hope this helps.
-
MonroePoteet
I do what you're talking about very regularly, doing the animation in Lightwave and producing TIF or TGA framestores, but I don't use the "pre-multiplied" alpha channels.
1) I don't think you can simply drag a sequence to the timeline. I always convert to UIS files, which works fine. The conversion doesn't take very long, and then MSP sees a UIS file as an animation.
2) The alpha channels are only available in V1 and above (not Va or Vb). You must explicitly turn them on by choosing the Overlay Options in the right-click menu on the clip, and then selecting Alpha Channel as the overlay type.
3) I think MSP is great for compositing, etc. I've done animation on Live video, multiple animations, etc. However, I've never used Premiere and After Effects, so I can't compare them.
mTp
1) I don't think you can simply drag a sequence to the timeline. I always convert to UIS files, which works fine. The conversion doesn't take very long, and then MSP sees a UIS file as an animation.
2) The alpha channels are only available in V1 and above (not Va or Vb). You must explicitly turn them on by choosing the Overlay Options in the right-click menu on the clip, and then selecting Alpha Channel as the overlay type.
3) I think MSP is great for compositing, etc. I've done animation on Live video, multiple animations, etc. However, I've never used Premiere and After Effects, so I can't compare them.
mTp
-
shawnm
Advantages of Premultiplied Alphas?
Thanks for your input. It sounds like you guys are using this software the way I want to. I am no expert at compositing, but at work we always used premultiplied alphas. I never really bothered to find out why. In fact, I don't really even know what "premultiplied alphas" are. What will I be giving up if I don't use premultiplied alphas? Do they allow cleaner edges? Or are they mostly used for more advanced compositing?
Thanks,
Shawn
Thanks,
Shawn
-
NickJushchyshyn
If you have a compositing application that supports both premultiplied and non-premultiplied, the only difference is the setting in your comp tool.
Here's the deal,
Say you've got image X to be composited over image Y.
X has an alpha channel, where 1.0 = solid, 0 = transparent, everything in between is some degree of partial transparency.
The traditional math for calculating any given color channel for any pixel is:
cX*alpha + cY *(1-alpha)
Virtually all compositing tools support this calculation, but it only works if the color channels in X have NOT been pre-multiplied by alpha. If they HAVE, then the math ends up being wrong, becaus the value stored in the color channel (cX) already equals cX*alpha.
If the traditional calcualtion is used (which is what MSP does) then the finished equation IMPLEMENTED is: (cX*alpha) * alpha + cY*(1-alpha). Sinc the partiall transparent pixels have been multiplied twice, it looks all wrong and you end up with an obvious edge to your comp.
So.... why not always render with non-premultiplied?
The thing about non premultiplied is that when you look at the images "raw" in a viewer that doesn't factor alpha, the edges look horrendous and "chunky". Like a render with no anti-aliasing switched on. This is because the edge pixels, that are supposed to be partially transparent, are showing as solid color.
Finally, what about compositing tools that DO support premultiplied alpha?
Well, they just modify the comp math, knowing that the alpha has already been applied to the color channels in that image X:
(pre-multiplied)cX + cY*(1-alpha)
So, if you're at work, where the pipeline is well established and it's known that all the comp tools support premultiplied alpha, the in-house production standard of rendering pre-multiplied works just fine.
If you're on your own, and have only tools that support the traditional math, you'll be better off rendering with non-premultiplied.
Hope this helps.
Have fun.
Here's the deal,
Say you've got image X to be composited over image Y.
X has an alpha channel, where 1.0 = solid, 0 = transparent, everything in between is some degree of partial transparency.
The traditional math for calculating any given color channel for any pixel is:
cX*alpha + cY *(1-alpha)
Virtually all compositing tools support this calculation, but it only works if the color channels in X have NOT been pre-multiplied by alpha. If they HAVE, then the math ends up being wrong, becaus the value stored in the color channel (cX) already equals cX*alpha.
If the traditional calcualtion is used (which is what MSP does) then the finished equation IMPLEMENTED is: (cX*alpha) * alpha + cY*(1-alpha). Sinc the partiall transparent pixels have been multiplied twice, it looks all wrong and you end up with an obvious edge to your comp.
So.... why not always render with non-premultiplied?
The thing about non premultiplied is that when you look at the images "raw" in a viewer that doesn't factor alpha, the edges look horrendous and "chunky". Like a render with no anti-aliasing switched on. This is because the edge pixels, that are supposed to be partially transparent, are showing as solid color.
Finally, what about compositing tools that DO support premultiplied alpha?
Well, they just modify the comp math, knowing that the alpha has already been applied to the color channels in that image X:
(pre-multiplied)cX + cY*(1-alpha)
So, if you're at work, where the pipeline is well established and it's known that all the comp tools support premultiplied alpha, the in-house production standard of rendering pre-multiplied works just fine.
If you're on your own, and have only tools that support the traditional math, you'll be better off rendering with non-premultiplied.
Hope this helps.
Have fun.