You are viewing an archive of the old fxhome.com forums. The community has since moved to hitfilm.com.

Can VisionLb render Straight Alpha instead of Premultiplied?

Posted: Sat, 20th Dec 2008, 3:35am

Post 1 of 12

Dufusyte

Force: 1000 | Joined: 16th Feb 2008 | Posts: 5

VisionLab User

Gold Member

It appears VisionLab renders out Premultiplied Alpha. I need it to render out Straight Alpha.

Is there a way for VisionLab to render out Straight Alpha instead of Premultiplied Alpha?
Posted: Tue, 28th Apr 2009, 3:37am

Post 2 of 12

Dufusyte

Force: 1000 | Joined: 16th Feb 2008 | Posts: 5

VisionLab User

Gold Member

Yes, no, maybe so?
Posted: Tue, 28th Apr 2009, 4:20am

Post 3 of 12

Axeman

Force: 17995 | Joined: 20th Jan 2002 | Posts: 6124

VisionLab User VideoWrap User PhotoKey 5 Pro User MuzzlePlug User PowerPlug User PhotoKey 3 Plug-in User FXhome Movie Maker FXpreset Maker MacOS User

SuperUser

I don't believe there is a setting for this in VisionLab, but how the alpha is handled should be controlled by whatever application you are using after VisionLab. What app are you using for your post-VisionLab compositing?
Posted: Tue, 28th Apr 2009, 5:30am

Post 4 of 12

spydurhank

Force: 1956 | Joined: 24th Jun 2008 | Posts: 1357

VisionLab User VideoWrap User FXpreset Maker Windows User FXhome Movie Maker

Gold Member

Are you asking if V.L. Can render out something with an alpha channel? Something that you first had to key out so that a certian part of your video clip will be transaparent when you place it into another app?

Like... say you have some grrenscreen footage, you place it into V.L. and key out the green right? So if you then place that same video clip into another N.L.E. or app, the part that used to be green is now transparent, So you can have whatever background behind the video clip?

If that is your question then please verify so we can tell you the proper way to do this.
Thanks. biggrin
Posted: Tue, 28th Apr 2009, 6:22am

Post 5 of 12

Axeman

Force: 17995 | Joined: 20th Jan 2002 | Posts: 6124

VisionLab User VideoWrap User PhotoKey 5 Pro User MuzzlePlug User PowerPlug User PhotoKey 3 Plug-in User FXhome Movie Maker FXpreset Maker MacOS User

SuperUser

What you are thinking of spydurhank is Premultiplied alpha. What Dufusyte is after, Straight Alpha, is just the black and white alpha channel, without the actual footage included. Just the mask, in other words.

In addition to what program you are bringing the footage into after VisionLab, if you can let us know why exactly you need the straight alpha we may be able to offer alternate suggestions to help you achieve whatever you are trying to achieve.
Posted: Tue, 28th Apr 2009, 8:33am

Post 6 of 12

spydurhank

Force: 1956 | Joined: 24th Jun 2008 | Posts: 1357

VisionLab User VideoWrap User FXpreset Maker Windows User FXhome Movie Maker

Gold Member

Axeman wrote:

What you are thinking of spydurhank is Premultiplied alpha. What Dufusyte is after, Straight Alpha, is just the black and white alpha channel, without the actual footage included. Just the mask, in other words.

In addition to what program you are bringing the footage into after VisionLab, if you can let us know why exactly you need the straight alpha we may be able to offer alternate suggestions to help you achieve whatever you are trying to achieve.
Aaah!
Okay, I see were you're going with this. That is so weird because I've actually been trying to figure out a simple way to do this in V.L. but I've been really hesitant to try anything out as of yet even though I really need to figure out how to to this.

Did you ever watch this?

http://www.youtube.com/watch?v=56-FQAxrCps

I did that frame by frame in Gimp. It took a long while.
Picture the same kind of effect, but the guy is wearing a sleevles shirt or vest. So rather than having to add the effect to each arm and then the head "frame by frame" in Gimp...

I was thinking of filming the transformation thing against a greenscreen.
Then stick the clip in V.L. and mask out only the vest and key out the green. Render it as an image stream and open it up in Gimp so that I'll only have to add the grade filter once on each individual frame rather than do it body part by body part in each single frame because everything but the body parts will be transparent.
Then using the original clip in V.L. with the garbage matte and chroma keyed info... simply invert the mask so that it'll only render the guy's cloths and nothing else. Then stick both clips together in V.L. and grade or color correct them so that it shows him before and after the transormation but his cloths remain the same.
We're going for a metallic, asphalt, concrete and ice look for now.
We're figuring on using practical makeup and applications for the rest. I may be thinking too much into it because I think I can do it but then I end up confusing myself.
I need to settle down and take it one step at a time.
Posted: Wed, 29th Apr 2009, 2:13am

Post 7 of 12

Dufusyte

Force: 1000 | Joined: 16th Feb 2008 | Posts: 5

VisionLab User

Gold Member

After keying in VisionLab (excellent keyer, btw), I'm bringing the footage into my NLE, Newtek's SpeedEdit.

As you know, there are different kinds of alpha (straight, black matte premultiplied, white matte premultipled, etc) and not every application can write every kind (for example VisionLab can only write Black Matte Premultiplied) and not every app can read every kind, for example SpeedEdit reads all alpha as Straight Alpha.

And when you have one kind of alpha coming from one app, and it is interpreted as a different kind of alpha in the other app, it tends to look funny around the edges, either getting a dark ring, or a light ring, or looking blobby, etc. So, after doing a great key in VisionLab, all the hard work gets misinterpreted, and it looks wonky in the next app.

For a good explanation of the different kinds of alpha, and what they look like when they are misinterpreted, see this article with pictures.

Now, I must say that Straight Alpha is actually the best kind, because it preserves the edge colors the widest, and so if you do a blur afterward, you are more likely to get good edge colors, whereas a Black Matte will tend to introduce blackness around the edges if you blur it, since the black is baked into the semi-transparent edges. Therefore, your higher end video production applications will tend to work in Straight Alpha, and sometimes exclusively Straight Alpha.

So, if FxHome has any aspirations to greatness (and the keyer really is top notch), it really needs an option to render out Straight Alpha. The article linked above even has sample code for alpha management.
Posted: Wed, 29th Apr 2009, 2:24am

Post 8 of 12

Axeman

Force: 17995 | Joined: 20th Jan 2002 | Posts: 6124

VisionLab User VideoWrap User PhotoKey 5 Pro User MuzzlePlug User PowerPlug User PhotoKey 3 Plug-in User FXhome Movie Maker FXpreset Maker MacOS User

SuperUser

Great link, thanks for that. Is there a reason you aren't just completing your composite in VisionLab? It seems like reducing the number of times you have to render is a good thing all around, although I do realize there are occasions when completing the composite in another app is necessary.
Posted: Wed, 29th Apr 2009, 3:16pm

Post 9 of 12

Dufusyte

Force: 1000 | Joined: 16th Feb 2008 | Posts: 5

VisionLab User

Gold Member

There are two roads we could take:
  • If fxhome improves their product, all its users benefit.
  • If I alter my workflow, I am inconvenienced.
Which road are you recommending?
Posted: Wed, 29th Apr 2009, 11:46pm

Post 10 of 12

Axeman

Force: 17995 | Joined: 20th Jan 2002 | Posts: 6124

VisionLab User VideoWrap User PhotoKey 5 Pro User MuzzlePlug User PowerPlug User PhotoKey 3 Plug-in User FXhome Movie Maker FXpreset Maker MacOS User

SuperUser

Oh, don't get me wrong, I'd love to see VisionLab have a wider array of alpha options as well. But I'm also fairly certain that its not something them can implement immediately. I was just trying to keep you going for the present, until such time as FXhome introduces the improvements you mentioned.

And without intending to insult you in any way, I was just curious as to the reasons behind your decision to composite in your NLE instead of your compositing program, as in most cases that seems like it would be a bit more awkward. I deal with a lot of users who are fairly new to these processes and aren't really clear on the best sequence of steps to accomplish what they are trying to do. I'm not saying you fall into this category, just that I find it best to ask, as sometimes I or other users can offer suggestions to improve their workflow, results, and efficiency.

So I was actually looking at road three, keeping you going with as little inconvenience as possible, while road one is under construction.
Posted: Fri, 1st May 2009, 1:44pm

Post 11 of 12

Dufusyte

Force: 1000 | Joined: 16th Feb 2008 | Posts: 5

VisionLab User

Gold Member

Axeman, thank you for your helpful suggestions. You really should win an award for being so helpful!

I'm actually more interested in all applications having proper alpha support than in my own insignificant project. I did invest in a compositing app which has excellent alpha support for input and output (one mentioned in the article linked above).

Projects often involve files coming from many sources and from many programs, and it is really essential to have a program that can correctly interpret the alpha coming in, and correctly output the kind of alpha that the next app may be expecting.

The bottom line about compositing is it is all about the alpha. If I were creating a compositing application, the first thing I would code is the alpha management. That is, the ability to correctly interpret the alpha of a source file, plus giving the user the ability to manually specify the alpha type of a source file if need be, as well as the ability to render out the type of alpha the user specifies.

The reason someone might not want to do the entire composite in VisionLab is the lack of motion tracking, splines, etc. But VisionLab's keyer is really excellent. So I find myself keying in VisionLab and then compositing in the next app.
Posted: Fri, 1st May 2009, 1:54pm

Post 12 of 12

Bolbi

Force: 408 | Joined: 22nd Apr 2006 | Posts: 429

EffectsLab Pro User MacOS User

Gold Member

Dufusyte wrote:

Axeman, thank you for your helpful suggestions. You really should win an award for being so helpful!
He has. wink