Although what you are asking for is a bit beyond what the Grain was designed for, I don't disagree that there should be an option to add a random noise generation in the addition of the grain. But just know this is an edge case. Ideally for motion you would prolly want to add a degree of control for the randomness so it behaves in a manner that would work better in motion (which again is NOT what ACR was designed for). Prolly want some sort of "jitter" that could be cycled so it looks nice, right?

This excellent exchange underscores the terrible deed Adobe committed when they bought Visual Infinity just to kill the superb set of plug-ins called Grain Surgery 2 (Add Grain, Match Grain, Remove Grain and Sample Grain), in my opinion the best film grain management tool ever (not digital noise), on any platform.

Adobe already owns this technology, they just shelved it.  It would be just a question of porting it over to work on current operating systems.  The last Mac version—and I don't recall off hand whether there was a Windows version—only runs on Power Macs and hosted by CS4 or lower, a resounding reason why I'll keep a G5 Quad running Tiger forever.

If Adobe chose to revive Grain Surgery and sell it separately under whatever name, I'd buy it in a heartbeat.  I thing they owe it to their shareholders. 

Richard, you, in turn now you don't seem to understand what is being asked for.

I'm not hearing Joe claim the Film Grain pattern being generated in any one image is unacceptable, nor (with all due respect to Jeff) do I see any reason to believe that patterns would emerge if it weren't kept constant from frame to frame.

Apparently the pseudo-random number generator that drives the noise added for the Film Grain feature starts with the same seed every time, meaning the pattern remains fixed frame after frame.  Joe just needs an option to have it start each time with a different seed.

Programmers strive to produce repeatable results, which is normally a Good Thing. 

But an option to "Increase Randomness" would be helpful in this case.  It would be easy to code - just avoid reinitialization of the seed to a fixed value - e.g., leave it alone or maybe use the clock (e.g., something that's constantly changing) to initialize it.

-Noel

Noel,

Can you grasp the difference between "I disagree with you, and here's why..." as opposed to "now you don't seem to understand what is being asked for"?

Why not just state your opinion without the personal references?

Richard Southworth

Eric, maybe you could use some kind of hash generated from the input data itself.  That would have the characteristic of always being consistent for that image, yet different for each image.  Just an idle thought.

-Noel

So you thought you were responding to me in kind, and I thought I was responding to your negativity.  Such is the way things can escalate on a public forum.

Tell you what, what say we both reset.  I really have no beef with you, and I do value your viewpoint on things.

I'll promise to be extra sensitive to hurting others' feelings if you will.  Deal?

-Noel

Joe_Mulleta wrote:

 

There is nothing to disagree Jeff, you obviously don't get the concept of "random"...

Random grain can't cause a pattern over time by definition. PERIOD.

Well, actually I do understand the concept of "random" or a stochastic process. A truly random generation of grain may not be really what you would want. A grain generation would pollly be better if it wasn't totally random but only appear random and produce a grain dance like you have with film which I'm not convinced is totally "random" but follows some patterns based on the coatings on film and the chemical processing of the film. The fact that the grain appears random doesn't mean it is 100% random. Add the additional factor of the movement caused by the film gate of the camera and projector you end up with what I call dancing grain because of the various factors. And yes, adding that sort of grain enhancement could be really cool.

But then the problem is how do you do that for the grain? Once you have the grain set, do you store that setting in XMP so it's consistent every time you process that same image? Or could you live with the grain being different every time you process that image? In the case of time lapse, it WOULD be useful to be able to apply a different grain pattern to sequential images…but as Eric has said, there are implications involved. The real need for a stochastic grain is an edge case.

I'm sure Eric will look into it, but how many users need it? Feature triage means that the engineering development do the greatest good for the greatest number of users on mission critical functionality. Of course, there are features that get added because they're, well, cool to do. Maybe stochastic grain and maybe even gate jitter may fall into this category...

Vit Novak wrote:

 

Guys, I think you are missing the point here

 

It wasn't the question about generation of grain pattern and how random it is, but about making grain pattern different in every frame

Agreed...but the question remains, can you live with a different grain generation for each time to process an image? If you want some sort of differences in the grain generation, what exactly are the implications for a single image setting? Does it change every time to touch the Grain setting? Do you store a specific setting in xmp so one can render the exact same rendering or do you change it every time? There are significant implications depending on what you want or expect...

Vit, but as you admit, what happens if you apply a given grain size to images whose sizes (or crops of the same size) are applied are different? Eric hinted at the potential issues involved...I think it's a harder problem that it appears...and is still an edge case at best. Again, I'm ok with looking at the issue up to a certain extent...but personally, dealing with individual images, adding a stochastic process to the grain isn't a high priority...if it can be done with a low pain threshold, fine...if it's gonna be a pain, then it's hard to argue for...