Fog made of a cube covering the whole scenery, with a simple volume scatter (not the best ? idk what’s the difference with the others…), at 0,01 density
Unfortunately, this specific use case is one of Cycles’ weakpoints because of noise. However, you can get around it and get a better result.
Rather than creating the light cone with an actual light interacting with the volume, you can instead create a separate shader which uses a node controlled emision volume to make an artificial light cone in a box which you place on each flashlight. You could look for an IES texture of a flashlight and use that as the basis of the effect to make it more realistic. For the 3D volume, check out the hogram tutorial by Bad Normals on youtube (I’d link it but I’m on mobile rn and it’s inconvenient). You should be able to just plug your texture into the node setup he makes for the projection effect, which is exactly what you need for your flashlight.
If the emission volumes cause their own noise problems, i think the light path node might be able to fix this but I’d need to do some testing and I’m not getting out of bed rn.
As for casting light on your scene, you could keep the spotlight and ise the IES texture in the light nodes to add detail. I think you might be able to make it exempt from the volume if you use the light path node’s “is diffuse ray” and “is glossy ray” outputs to still light your scene without interacting with the volume, but again i’m not sure how it would behave and I’m not going to test it rn.
Hopefully something in here is helpful. I’m really curious how my idea for a fix will work since I haven’t done it that exact way before, so if there’s issues I’d be happy to do some testing myself tomorrow
Thank you for your response, but, after testing out your solutions, It was looking cleaner, but I loosed the atmosphere feeling I wanted to have.
So I guess that, as I’m using it for a 2d still image, I’ll do a mixtup of both of the images to try and get a good final result
Have you tried denoising the volume separately from the rest of the render? It can give a cleaner result.
There are 2 ways to do it.
Render and denoise the volume as a separate render from the rest of the scene, then composite it back in. You first render the scene without the volume. Then, you do a second render where every object except the volume has been set to holdout. You give the render a transparent backround (the option is in the “film” section of the render setting, plus you also set the output format to rgba mode, so it has transparency). That will give you a second render with only the volume on a transparent background.
The SID denoiser addon could help you here. It setups the compositor so every render pass (diffuse, glossy, transparent, volume) get denoised separately from each other and then recombined, which can make the denoising look more detailed and clean. With the more recent versions of the denoiser, this isn’t as useful as it once was, but it could still be useful for volumes. It’s at least worth trying.
Also, Have you tried the different volume sampling types? Depending on the scene, it won’t always be the same that gives the best result.
If you use a volume scatter alone, it will redirect the light, but it won’t simulate absorption. The light won’t get blocked and get reduced as it travels through the volume, so it will act more like a transparent gas than like dust. Normally, volume scatter and absorption are meant to be used together. The easiest way to do it is to use the principled volume shader: it contains every aspect of a volume in a single node.
But if your current volume looks good artistically, that’s perfectly fine. Just be aware it’s not fully respecting the real life physics of a dusty atmosphere.
Actually you should try opposite. For “pure” cycles solution, try to reduce bounces to minimum where you can see difference, something like 3 diffuse, 1 glossy and so on, and add incrementally until there is no improvement. Also you should try to disable caustics if they doesn’t make much difference. Using default volume steps and step rate is good to start and that is for improving volume accuracy in depth axis.
Then, try to add samples high, something like 65536 and and set adaptive noise threshold to 0.0025. You can then try to adjust amount of noise you prefer to image. For noisier image you can use 0.1.