Fire flies / noise help...

[tris3d]

Hello,

I'm a 3ds Max / Vray user and I've been following the progress of Fstorm render for a while now (in the background) and thought I would give it a go for a client Christmas animation project. I've stumbled on to my first problem, I can't seem to avoid the fireflies / noise in the image attached. I searched the forum looking for information, I thought reducing the GI clamp might work but it didn't effect it. After my failed search I decided to just get into the settings and literally play with every parameter, cranking the settings high one by one to eliminate the issue. Unfortunately, that hasn't worked either.

Now, I could just admit defeat and move the baubles that are causing the issue but I really want to learn so that I can have solutions if faced with a similar issue in the future.

Please don't judge the quality of the image lol, I've literally just started playing with the scene so it's a little embarrassing sharing.

Any help or a pointer in the right direction would really help.

Many thanks,

Tristan

[antonyebl]

That'd be DOF effect you've highlighted. (there's a DOF override on the camera in your scene;
also christmas ball materials have light emission on for some reason btw)
Maybe with more samples (> 1500) noise would clear) or lower dof radius.

[tris3d]

Hey Anton! Thanks a lot for your reply, really appreciate it!

Yeah, it's definitely to do with the dof, I should have mentioned that. I did try upping the samples! In fact, I upped it to over 5000 to rule out that setting.

In terms of lowering the dof, I'm aiming for that particular look and it would be a shame to substitute creative decisions due to render issues.

Can you think of any other solutions? Sorry to be a pain! I can upload the scene tomorrow if that would help?

[antonyebl]

To start with I'd switch off the emissivity of the material (to avoid light source hotspots/fireflies);
then render the most affected regions with more samples with 'region render'

I don't have your HDRI so the result isn't the same but with ~22k samples it looks ok to me (based on the noise adaptivity it still indicates only 4.5% noise free area so perhaps even more could do)

[Javadevil]

I don't think render region is a good solution here, the Fstorm philosophy is its strength in Adaptivity, Its MIS, knowing where to sample more.

You could try raising the minimum samples.

Otherwise I think this is a scene for Karba to take a look at.

[antonyebl]

Well yes but from my quick test the image looked pretty good @1000 samples except those highlighted DOF ares;
to speed things up (as I have some other stuff to finish I set a render region to apply adaptivity within that small area and there was still noise at 22k samples.

I believe the adaptivity would eventually transpire across the whole image equally, but this was faster...

[tris3d]

Apologies, I should have included the hdri for more accurate testing. Really appreciate you having a look at this for me. I'm pretty sure I did up the samples to 20,000 odd when testing which is strange!

I didnt even even realise I had emissivity turned on, to be honest, it was a vray material conversion. Maybe this has a larger effect than the samples. I'll have a look first thing tomorrow at work and let you know!

Again, I really appreciate the help with this! It would have been so easy to switch back to vray as obviously deadlines loom. For the sake of learning and adapting a new workflow, I really want to stick with fstorm.