"Multiply" render error (5.6 + 6.2)
Moderators: Víctor Paredes, Belgarath, slowtiger
"Multiply" render error (5.6 + 6.2)
http://slowtiger.de/examples/cels.mov
This strange render is the result of duplicating one layer which is set to "multiply" (then scaling it). It's only 2 layers inside a bone layer, animated with the "snake" trick.
Adding more "multiply" layers results in everything rendered black. Any idea how to overcome this? I've never had this problem before.
The "cels" are vector shapes with some halo inside.
UPDATE: I tried the same with an image layer (cels on transparent): same result, the render seems to use the alpha channel only and looses all colour information.
This strange render is the result of duplicating one layer which is set to "multiply" (then scaling it). It's only 2 layers inside a bone layer, animated with the "snake" trick.
Adding more "multiply" layers results in everything rendered black. Any idea how to overcome this? I've never had this problem before.
The "cels" are vector shapes with some halo inside.
UPDATE: I tried the same with an image layer (cels on transparent): same result, the render seems to use the alpha channel only and looses all colour information.
I built again everything from scratch this time, and the result until here was fine:
http://slowtiger.de/examples/mu_test4.mov
The setup is quite elaborate. It consists of a cel vector layer, where each cel has a half-transparent fill, a halo, and a blurred outline.
- top group layer
- - bone layer
- - - cel vector layer
- - - cel vector layer
- - bone layer
- - - cel vector layer
- - - cel vector layer
- - - cel vector layer
The fun started when I set layers to transparent. This setup still worked:
- top group layer 100%
- - bone layer 100%
- - - cel vector layer 80%
- - - cel vector layer 80%
- - bone layer 100%
- - - cel vector layer 100%
- - - cel vector layer 100%
- - - cel vector layer 80%
and looks like the link above.
This setup doesn't work (the only difference highlighted):
- top group layer 80%
- - bone layer 100%
- - - cel vector layer 80%
- - - cel vector layer 80%
- - bone layer 100%
- - - cel vector layer 100%
- - - cel vector layer 100%
- - - cel vector layer 80%
and looks like this: http://slowtiger.de/examples/mu_test4a.mov. All colour information is lost.
This setup doesn't work in a different way:
- top group layer 100%
- - bone layer 80%
- - - cel vector layer 80%
- - - cel vector layer 80%
- - bone layer 100%
- - - cel vector layer 100%
- - - cel vector layer 100%
- - - cel vector layer 80%
and resulted in only one layer being gray.
I think this counts as a bug, as AS 5.6 and 6.2 seem to hava a limit with nested transparencies. I'd like to hear if others can confirm this.
Have a try yourself with the file, http://slowtiger.de/examples/mu_test4a.anme.zip.
http://slowtiger.de/examples/mu_test4.mov
The setup is quite elaborate. It consists of a cel vector layer, where each cel has a half-transparent fill, a halo, and a blurred outline.
- top group layer
- - bone layer
- - - cel vector layer
- - - cel vector layer
- - bone layer
- - - cel vector layer
- - - cel vector layer
- - - cel vector layer
The fun started when I set layers to transparent. This setup still worked:
- top group layer 100%
- - bone layer 100%
- - - cel vector layer 80%
- - - cel vector layer 80%
- - bone layer 100%
- - - cel vector layer 100%
- - - cel vector layer 100%
- - - cel vector layer 80%
and looks like the link above.
This setup doesn't work (the only difference highlighted):
- top group layer 80%
- - bone layer 100%
- - - cel vector layer 80%
- - - cel vector layer 80%
- - bone layer 100%
- - - cel vector layer 100%
- - - cel vector layer 100%
- - - cel vector layer 80%
and looks like this: http://slowtiger.de/examples/mu_test4a.mov. All colour information is lost.
This setup doesn't work in a different way:
- top group layer 100%
- - bone layer 80%
- - - cel vector layer 80%
- - - cel vector layer 80%
- - bone layer 100%
- - - cel vector layer 100%
- - - cel vector layer 100%
- - - cel vector layer 80%
and resulted in only one layer being gray.
I think this counts as a bug, as AS 5.6 and 6.2 seem to hava a limit with nested transparencies. I'd like to hear if others can confirm this.
Have a try yourself with the file, http://slowtiger.de/examples/mu_test4a.anme.zip.
Ink effects like multiply have "weird" results when there is no background. If there is nothing behind the layers with an ink effect it doesn't look right (grayscale or black). I've discussed this with Mike already but it's been a while. He did explain why it was an issue and tricky to fix. Photoshop sort of does the same thing when you use ink effects against transparency but not quite. Ink effects like multiply need to "multiply" the pixel values against other pixel values. I guess if there are no pixel values to multiply weird things happen.
Put a single vector layer with a white filled shape on it behind all layers in your file then render. The color comes back. I think the problem is the nesting and the top layer with an ink effect that causes it to go "grayscale" because it has nothing to "multiply" against. Your other tests worked because of other layers lower in the nesting? Just a guess.
-vern
Put a single vector layer with a white filled shape on it behind all layers in your file then render. The color comes back. I think the problem is the nesting and the top layer with an ink effect that causes it to go "grayscale" because it has nothing to "multiply" against. Your other tests worked because of other layers lower in the nesting? Just a guess.
-vern
- Víctor Paredes
- Site Admin
- Posts: 5818
- Joined: Wed Jan 26, 2005 12:18 am
- Location: Barcelona/Chile
- Contact:
I have had very weird multiply results some time ago on version 7.
I forgot to post it here, I'll search if I can find the file.
I forgot to post it here, I'll search if I can find the file.






Moho co-owner
Previously Rigged animation supervisor: My father's dragon, Wolfwalkers & Star Wars Visions "Screecher's Reach"
My personal Youtube Channel
If your assumption were right, then the b/w effect would be visible over the transparent BG only, but not where shape intersects another shape. Also the bottom layers should be affected first, which isn't the case:
http://slowtiger.de/examples/mu_test4b.mov
Anyway, I don't think this is going to be fixed soon, so I just look out for this bug where it could happen, and work around it.
http://slowtiger.de/examples/mu_test4b.mov
Anyway, I don't think this is going to be fixed soon, so I just look out for this bug where it could happen, and work around it.
Slowtiger,
Yes, sort of right except that all the layers are "multiply" so they would all have the same issue against a transparent background. For example if one layer is grayscale against transparent then another layer above it will just be multiplied against grayscale.
I did discuss this with mike as a bug. As I recall he either didn't agree 100% or it might be he didn't have an easy solution for it. Hard to remember. I recall that the main issue is what to do with the pixels values when there was nothing to render against. I had the same problem with a file. If you don't use transparency on the layer an ink effect set to multiply just appears totally black.
-vern
Yes, sort of right except that all the layers are "multiply" so they would all have the same issue against a transparent background. For example if one layer is grayscale against transparent then another layer above it will just be multiplied against grayscale.
I did discuss this with mike as a bug. As I recall he either didn't agree 100% or it might be he didn't have an easy solution for it. Hard to remember. I recall that the main issue is what to do with the pixels values when there was nothing to render against. I had the same problem with a file. If you don't use transparency on the layer an ink effect set to multiply just appears totally black.
-vern
- funksmaname
- Posts: 3174
- Joined: Tue May 29, 2007 11:31 am
- Location: New Zealand
- funksmaname
- Posts: 3174
- Joined: Tue May 29, 2007 11:31 am
- Location: New Zealand
true, but you could render once with the background layer (colour), and once without (effectively a luma matte), and then use the grayscale version to mask in post using the other programs... or you could do an alpha pass without the background (might be more accurate than the grayscale) and mix that with a 'solid' export?
I realise it's not ideal though...
I realise it's not ideal though...