Chatgpt4 + Persistence = DasGrain for Flame, maybe?

Flame Matchbox
1

We’ve all been there or will be at some point, the battle of grain. Most of the time the good old plate, denoised plate, and two comp nodes will set ya straight, but not all the time. For me, I had to blow up a shot 200% and add a fake pan, but that pesky high-frequency detail revealed a ghost of what once was. No amount of front matte offsets on the noise pass could save me. This shot is for long-form, and the Regrain node won’t cut it. I had to throw in the towel and have a Nuke Artist kick me out of a grain pass with DasGrain. Just like that, the battle was won, but the war was not over.

Is it possible to replicate DasGrain’s normalized grain pass in Flame?

Based, on my short research the problem we Flamers need to solve is DasGran’s Scatter tool. So far, I think I’ve figured out the area plot:

  • Connect the Motif node to the noise pass
  • Hit F1
  • Change the ROI to square
  • Select an area of noise without any detail in it
  • Hit F4 to view

Others have used the 2dtransform to get here. Both work. Which is better? I have no idea.

Okay, we have an area of noise without detail spread throughout the entire frame, but the grid pattern is extremely noticeable. How can we solve this? I believe DasGran solves this through the use of a Voronoi pattern. This past week I’ve been using Chatgpt4 to write a matchbox shader that will separate and scramble an input using a Voronoi pattern, but I think I’ve hit my wall which has led me to write this post and reach out to the community.

Is it possible for a matchbox shader to separate and scramble an input through the use of the Voronoi pattern/function? More than happy to share my Chatgpt4 matchbox shader workflow if that’s the case.

9 Likes
2

Ivar’s Crok_Seemless set to Voronoi for the win?

image
image726×416 39.6 KB

6 Likes
3

Das Grain can`t be replicated with a matchbox shader. The problem lies not in a Voronoi scatter (can be easy to do with Action + right UVmap). Das Grain synthesizes patch grain using several frames which is not possible.

PS If I am not mistaken crok_seamless uses filtering, this is no go when replicating a grain. I could be wrong.

2 Likes
4

I saw your post in the regrain thread but couldn’t wrap my head around it. Any chance you could share a pic of your normalized grain in flame setup?

1 Like
5

I dont have access to Flame now((( Kyiv, Ukraine, war, all these little problems. I can share a Fusion setup for reverse engineering, or, let me ask Andy if its possible to remotely access someone`s Flame to recreate it.

1 Like
6

I’ll see if I can setup something in a few days… although latency and bandwidth might make it impossible.

1 Like
7

Thanks, Alan. Teamviewer or screen share with someone on the other end will do the job.

2 Likes
8

We only use Teradici here, but the client is free and multi-platform.

9

Apologies, I forgot about your situation. I appreciate your engagement and will give your comment in the other thread another shot. Thank you! :fire:

10

check your chat.

11

i really hope that someone at adsk hears us screaming for a nee regrain tool

dasgrain is pretty much in every comp now, it became as oblivious as neatvideo.

4 Likes
12

@val I can give you access to a Flame as well if need be via the Teradici PCoIP Software Client and Tailscale, both free to you.

13

@randy @val

Randy & Val, can you give it a shot first, and if that doesn’t work out, I’ll be back up.

2 Likes
14

Word. Val is in.

4 Likes
15

They know, they no.

16

Yeah, I agree. Vote for this if you haven’t already.

2 Likes
17

unfortunately, this shader soft edge blends the cells.

2 Likes
18

Up Voted.

Open question. This das grain gizmo seems to be written in Python. Would it be possible to reverse engineer/translate and implement within Pybox?

1 Like
19

@AustinCampbell i wonder if you could open up Ivar’s shader and see if that could be used as a base by eliminating the softness in the stitching.

2 Likes
20

@randy Good idea! I’ll give that shot sometime this week.

4 Likes