Regrain in ACES

Sorry if this is easy, but I am not getting it.
Most (99%) of my work in TVC at REC709. No linear workflow stuff.
My regular clean up goes like this: Neat video to take the grain/noise out, do the clean up work, and then re-apply the noise via subtract-add comp nodes. Works most of the time no problem.
Working now on an ACES job, and the adding the grain back this way just does not give me the results I expect. Does this have something to do with the way linear video behaves?
Any ideas would be helpful, as I am loosing whats left of my hair just trying to add grain back.
Thanks.

1 Like

I try not to re grain in ACES (linear) but that isn’t a good answer.

Can you denoise, subtract from the orignal plate and then add it back to the denoise plate and make it look the same?
The order of subtraction is important. Which one is subtracted from the other matters.

I just did a test and it seemed to work for me in ACES. What type of error are you getting?

For linear workflow, try dividing de the original plate by the degrained one. Then multiply the result by the comp. Depending on the comp work done, the regrain may not be even.

1 Like

Yeah something ain’t right. 2-1 =1 in ACES.

1 Like

The add/subtract grain trick works best if your footage is log.

My workflow:

  1. Convert plate from ACES (or ACEScg, or Scene Linear Alexa Wide Gamut, Etc) to ACEScct via the Input transform of a Color Management node
  2. Denoise the plate
  3. Subtract the denoised plate from the original plate
  4. Convert the denoised plate back to ACEScg with another Input transform (copy the first one and hit the invert button)
  5. Comp the shot in linear.
  6. Convert the comped shot BACK to ACEScct again
  7. Add the grain subtract pass back on
  8. Convert the shot to ACEScg one final time and render.

I did a quick test and the attached image shows the difference. In the linear image, the altered area gets mustardy and very little grain is added back in spots. The log image shows the usual artifacts where details were altered, but the grain is more consistent and does not have any color shifts.

6 Likes

If doing the subtract method - I will subtract and add grain in linear (whatever flavour).

If graining back where you can’t add the original grain, I have used Log to add grain back A LOT am a strong advocate for this. However, my head has been turned recently by Crok Renoise - graining in Linear as it’s got a nice Alexa grain and you can change grain size, amount and there are curves. It’s pretty decent. Haven’t done any Episodic or cinema stuff with it though - just Adverts at 2 to 6k.

3 Likes

dasGrain… I just drag the thing though nuke and export a grain plate and then import back into flame … voila :slight_smile:

4 Likes

That shit should be pyboxed tout suite.

3 Likes

for real!

1 Like

Tried to follow along here, but when i convert the comp back, but when I switch the color management node to input transform, flame shows no result.

Hard to say without looking at the setup. The usual culprit is the input transform defaulting to “from rules”. If the comp isn’t tagged correctly (and there are plenty of ways that can occur) those rules don’t know what to do.

Specify what the incoming and outgoing colorspaces are it should work.

Given that it’s just (italics to indicate ignorance/sarcasm) a Nuke GIzmo, shouldn’t we be able to open up said Gizmo and re-create the work it’s doing in a flame context? And if we can rebuild it in flame AND the creator is okay with us doing so, couldn’t we have it in flame?

Ivar tried to recreate it as a matchbox… it should be doable I mean its all
open source so … try? :smiley: dasGrain rulez.

Many have tried. None have succeeded.

I may give it a go one day soon.

1 Like

No result as in black frame? Or no result as in No Result in the brim right corner? If so it’s likely a bit depth error. Neat outputs 32bit.

@andy_dill - i gave it a half-assed try about 2 years ago while chatting with Ivar about it. He said it was a frequent request. i agree with @finnjaeger it would be amazing to have in flame. Maybe @kyleobley has made some headway here…

1 Like

Sorry to disappoint Tim, but I’ve never really put much time into it. If Ivar wasn’t able to crack that nut there’s no chance I’ll be able to do it!

1 Like