Hi there!
Got this here as well. Solution is quite easy: just add a FrameHold node before
the Grain input of F_Regrain, so every frame grabs the grain from the same
source image.
But yes, something like the Lock Noise Analysis button in the Denoise node
would also be very helpful in the
...@beingfrank.info
To: nuke-users@support.thefoundry.co.uk
Sent: Friday, 9 August 2013, 4:20
Subject: Re: [Nuke-users] F_Regrain unreliable on render farm?
thanks guys.
yes, that's how I have been generating grain plates, but I found I got
better results using F_Regrain live - when and if it works
Hi all,
this is reported but was wondering if I'm the only one who finds
F_Regrain close to unusable in a render farm environment?
It's all dandy as long as you render locally, but as soon as you have
multiple machines render a script with F_Regrain in it, some of the
frame batches are all
I always put a frame hold after the plate going into the grain sample and it is
reliable on a render farm. Without the frame hold it will sample every frame
or batch I think.
John Mangia
Sent from my iPad
On Aug 8, 2013, at 10:48 PM, Frank Rueter fr...@beingfrank.info wrote:
Hi all,
I've always found the F_Regrain unreliable, for various reasons:
* it used to redo the analysis on each frame when on the farm, so you
had to HoldFrame the grain-src input. This may have been fixed, but I
still do this out of paranoia
* sometimes, the grain seed seems to change - e.g when
thanks guys.
yes, that's how I have been generating grain plates, but I found I got
better results using F_Regrain live - when and if it works.
cheers,
frank
On 09/08/13 15:12, Ben Dickson wrote:
I've always found the F_Regrain unreliable, for various reasons:
* it used to redo the
I think you lose the F_Regrain's grain response stuff.
Not used it recently, but I seem to recall getting almost exact matches
if I multiplied the prerendered-grain plate by specific values..
If not I guess it either does some clever magic with the response curve,
or just applies the grain
yup, been playing with multiplying it to match the live output better
and I'm getting there, at least close enough. The response can always be
fudged with some luma keys ;)
Certainly better than the headache F_Regrain causes.
Cheers,
frank
On 09/08/13 15:34, Ben Dickson wrote:
I think you