Fstorm render

[dean_dmoo]

I’ll give it a go then…

A very brief impression of my first 5min experience to see what all the fuss is about.

Tried Fstorm earlier. Ive never really bothered with RT as the scenes I work on are nearly always too big anyway. So for me they are pretty much the same in that respect. Anyhow, opened the chicago loft by BBB, converted and hit render. Didnt work straight away, but a couple of edits got it moving. Huge glare, dramatic image. What I liked was that I could see what was happening, mostly down to the bar at the top of the renderframe, counter for samples, remaining time to cleanup, noise value. Even though initially not much was happening on screen you could see it was doing something, even when my renderframe was staying black.

Tried to change some materials but crashed out of max twice. So opened the scene again to use vrayRT to see how it compares. Scene rendered how it should and I could see file loading and LC progress but once it started rendering there was no feedback
besides the actual image cleaning up. Now had to zoom in and really look to see that it was actually doing anything as at one point I reallu wondered whether it has stopped working.

For me, as pretty much a new user for both, that was the biggest UI difference I saw and what I liked with FStorm. That you could see a countdown of some kind whilst rendering, whereas with RT it was blank. Maybe it was down to settings that I couldnt see that, so let me know. But as a user, I like seeing that kind of feedback, similar to how you see passes in progressive.

[vlado]

@ttdww

For me, as pretty much a new user for both, that was the biggest UI difference I saw and what I liked with FStorm. That you could see a countdown of some kind whilst rendering, whereas with RT it was blank. Maybe it was down to settings that I couldnt see that, so let me know. But as a user, I like seeing that kind of feedback, similar to how you see passes in progressive.

I don’t know if it’s good or bad that it’s the only difference There is an option in V-Ray RT called “Show statistics” that would overlay some information in the VFB, like noise level reached, GPU load etc - see attachments.

Best regards,
Vlado

[peteristrate]

@joa_grilo

I don’t know if it’s good or bad that it’s the only difference There is an option in V-Ray RT called “Show statistics” that would overlay some information in the VFB, like noise level reached, GPU load etc - see attachments.
Best regards,
Vlado

I think we’re getting down to many people not knowing how to use VrayRT (or even VrayAdvanced) properly to get the most of it
I’m super happy with how the VrayRT development goes, given that many features will be implemented as well.
All I was looking for was some sort of a more realistic tonemapping, although not sure if that is the real issue, as we all discussed here.

[vlado]

@STACCATO_3D

All I was looking for was some sort of a more realistic tonemapping…

Yes, will see about that…

Best regards,
Vlado

[dean_dmoo]

Thanks Vlado!

@ Peter,

That was the point I was trying to make. When people post how “easy” Fstorm is I was curious if that was that way if you view it as a beginner (which for RT I am). As opposed to quite a few here who have years of experience and go very deep into the analysis. It can been interesting to take a look from a different perspective to find out which parts are actually making it seem that way and whether it is worthwhile for Vlado and the team to take it onboard. Is it down to tonemapping? Or down to how noise gets cleaned? Or lack of menu’s because it’s still a very young piece of software? The way they handle Glare?

I actually turned the statistics off, right at the beginning when the geometry was loading (that Vlado highlights) because I find it ugly having to view the image with the text. Not realising that that also turned off the progress being shown. So for that part my conclusion would be whether it is possible to have that info placed somewhere else on screen to keep your actual image clean. Just my 2cent’s. Something very small, but it does effect how “technical” the process gets shown by the renderer and something that also has an effect how people experience how “easy” it is.

[vlado]

@tashko.zashev

That was the point I was trying to make. When people post how “easy” Fstorm is I was curious if that was that way if you view it as a beginner (which for RT I am).

I guess it all depends on your viewpoint. For example if, for whatever reason, you have to produce a linear image out of FStorm, do you know which three or four options you need to change in order to get it? It should be an “easy” task… (That’s outside the fact that for the moment FStorm can’t render out an HDR image with the 3ds Max VFB - but that’s something that Karba has simply overlooked for the moment.)

it is possible to have that info placed somewhere else on screen to keep your actual image clean. Just my 2cent’s. Something very small, but it does effect how “technical” the process gets shown by the renderer and something that also has an effect how people experience how “easy” it is.That’s what we can do with the current VFB implementation right now. We have ideas for expanding the VFB to include extra information, but it’s not there yet.

Best regards,
Vlado

[snivlem]

Never tried FStorm due to being on Maya (but I have used octane/corona etc..) but over this last year VRayRT has made huge leaps and bounds I must say. It went from being a novelty to something it is constantly used, no longer do I need a “draft settings” script so I could test materials, I just now fire up RT, tweak and tweak and it’s great.

It looks like there is an illusion of these new renderers in that they seem more photorealistic but in reality it’s because they are locked down, don’t have as many features and just apply a really nice tonemapper and are optimised for small indoor scenes. I do like the unbiased lighting but I’m not sure how much of a discrepancy there is between using a Brute force workflow and some of these unbiased engines in a real world project…especially an exterior render.

I think if V-Ray got a modern filmic tonemapper and better integration of lens effects (subjective I know, but I can’t get a nice result usually compared to arionfx or the like) there wouldn’t be any difference apart from that you can do with more in V-Ray if you want the flexibility.

It reminds me of people using old proven full frame DSLR that are perfect for work yet Sony come out and release a mirror-less full frame camera changes the game and supposedly makes you a better photographer and people buy in ditching their perfectly good and sometimes better DSLR. In the end using a new renderer might be faster and better at particular things but it won’t make you a better artist.

[kosso_olli]

Here is my take on fStorm and I have to say that I don’t understand what all the fuss is about…

An overview of what I have done: I opened an old interior scene I had on my disk and converted it both for RT GPU and fStorm. For fStorm I disabled all the tone mapping, contrast, vignetting and stuff, because I wanted to see it’s raw result. You can see both images in grey below, and you can see that fStorm has a little more contrast. It doesn’t look better or more realistic, it doesn’t look worse.
Granted, fStorm looks a bit cleaner for the same amount of time. Some artifacts on the ceiling next to the cutouts though.

With materials however, that’s a different story. RT GPU looks much cleaner for the same amount of time (around 9mins for both). Also, something seems wrong with the materials in fStorm. I mean, the pillows look almost identical, but all the rest seems off. I am pretty sure that RT GPU would be way ahead with Light Cache as secondary ray, it was BF/BF in this case. Also, the Vray scene had two sided materials for the lamps
Also, interactivity is the same for both in RT mode. fStorm is not faster in that regard.

Greyshade:

RGB:

[re_FORM]

Nice test. Can you check the gamma on your reflection maps? It looks like the fstorm render has either stronger or sharper reflections from the window (more contrasty reflection map maybe?) You can also see a gamma shift on the wallpaper behind the shelfes, and the bed cover. I assume these changes are the result of some gamma funk rather than the renderer settings ? (eg reflection bounces?)

[kosso_olli]

@dp_mg42

You can also see a gamma shift on the wallpaper behind the shelfes, and the bed cover. I assume these changes are the result of some gamma funk rather than the renderer settings ? (eg reflection bounces?)

Thats what I see think, too. Didn’t find a way to fix this. Anyway, this was just a test to see what fStorm does better. Not so much, apart from Tonemapping and Glare, when enabled. If you don’t use these features, I see no point in using it.

[_Lele]

@bdancer

Here is my take on fStorm and I have to say that I don’t understand what all the fuss is about…

[joconnell]

That’s kind of it though, Oliver disabled all the “tone mapping, contrast, vignetting and stuff” to make it more like a standard linear renderer (vray’s default). Seems like the “photographicness” that people are liking about the default output of fstorm is all of this stuff which you can add in post if you want to, they’re just assuming that a lot of it is the way fstorm is calculating it’s lighting and shading instead of just colour correction bits after. Of course with me passing on all my renders to a comper I absolutely do NOT want any of this stuff on, but if you’re looking for an image right out of the box it’s closer to what would have come out of a dslr or film camera since both of them would have the tone mapping and canon / nikon “look” burnt into the image.

Might be worth expanding on your vray luts and doing a small video on how you’d achieve this in vray lele? Even something like making a set of presets that can get loaded into the globals button of the frame buffer to mimic fstorm’s response?

If people want non linear images saved from the frame buffer and that suits their business model then why not let them have it

[Morne]

@stevesideas

Might be worth expanding on your vray luts and doing a small video on how you’d achieve this in vray lele? Even something like making a set of presets that can get loaded into the globals button of the frame buffer to mimic fstorm’s response?

+1 for this Lele. Your vids are always very informative and fun

[joconnell]

Might also be worth putting in a vignetting option into the buffer cc controls!

[mitviz]

i think that was the point, not to strip down and render and try to compare them with similar settings just as is out of the box, most people do not pass images on to anyone else to work on, well the ones i know anyway and just want the most realistic images possible and with the least amount of work so its what fstorm offers not comparing standard linear renders, and also i don’t really think anyone is making a fuss about anything really, gpu rendering is not something alot of people are tackling right now anyway and is kind of new to alot of people, i can count the amount of people i know with a titan card and its about 1 and i know alot of people in my city and back home and they are all using vray and rendering so i think alot of people like myself learnt alot so far from this thread though. if very had say an option to do with fstorm does out of the box with gpu that would be really nice too

[joconnell]

@Morbid_Angel

Most people do not pass images on to anyone else to work on, well the ones i know anyway and just want the most realistic images possible and with the least amount of work so its what fstorm offers not comparing standard linear renders.

I’d say a huge percent of people pass on images to another point to be worked on, it’s just far quicker to do a lot of things in 2d than to do a full re-render in 3d. Your point on some people wanting “the most realistic images possible and with the least amount of work” is totally valid though. Fstorm’s defaults are making decisions on behalf of the artist and what’d happen is while I might like the look of the render, what it does to it would totally break my workflow by taking everything away from linear.

It’s just the issue with vray being a general purpose renderer - you won’t find one set of defaults or workflow that’ll suit every user or industries needs

[_Lele]

My attempts at figuring fStorm’s internal CM out have proven fruitless (without CM active, i get seriously wonky LUTs out of it), while Vlado has made some inroad, apparently (see posts earlier).
Notice that the same method, used with V-Ray LWF and with fStorm’s CMd output, produced valid LUTs, and only V-Ray’s was identical to the control (ie. straight from the CMSPattern) one.
fStorms skewed it a wee bit with CM on, and killed it (again, just in my tests: Vlado managed something, going down another route.) when trying to render in LWF.

fStorm installs all its luts into the max root, under the LUT folder, so nothing stops people from loading those or, fearing copyright issues, reading the names and finding equivalent ones, but royalty free, online.

There is no such a thing as the Nikon or Canon look: people shoot the cleanest raw they possibly can, and choose cameras based on the inherent quality of their response, these days only differing between makers because of they way each goes about sensors, surely not because of a post tone operation :ie. maybe the primaries are slightly different in gamut shape, that’s all. They ALL, unequivocally, aim at the closest possible RAW capture of light (AKA, in Rendering: LWF), and what stops them from achieving it is physics and technology (ie. it’s a limit, not a choice.).

All the LUTs coming with fStorm are for old stock film, and the overwhelming majority has to be applied in Log space.
NONE works out of the box, and that’s why they add a slider to blend them back: it’s instagramming, and done wrong at that.
The fact one doesn’t grasp what’s going wrong (f.e. using a LUT on an image which doesn’t reflect the CM the LUT was built upon, or applying a LUT in the wrong color space.) isn’t cured by a blend slider, and it’s genuinely not a behaviour I feel i can encourage (I scripted the LUT explorer begrudging the choice before even starting.).

If someone just once came up with actionable data, with a workflow showing the differences in a truly qualitative way (“I like this more” isn’t quite cutting it. Nor is posting overcontrasted stuff with severely cut off GI, and calling it “real”.), then maybe there could be a debate on what V-Ray ought to do better.

The ONLY thing so far of note is that V-Ray renders the correct lighting distribution, unerringly (like, say, Arnold, renderman, mental ray, Corona), while fStorm does not, repeatedly so, across all its releases so far (like, say, Octane, redshift, and most other GPU engines out there).
Now, that bit is exactly measurable, and if it was off, we’d have to fix it.
Or, we could tell you that it looks more photographic, turn glare on by default, and add the intensity the GI engine can’t capture back as an overlay that way.
Notice this isn’t a go at Karba: the guy’s doing great, he’s doing it alone, and I wouldn’t know where to begin from at the same task.
It’s expression of confusion at the priorities in people’s minds, as i seem to gather from this thread, and others of similar tone: they mostly leave me at a loss for words, and even more so of productive ideas.
Luckily, me alone.

[kosso_olli]

@Ymauchi

it’s not about what experienced people get out of Fstorm, but what inexperienced or not technically oriented people get out of it.

Take a look at the fStorm Facebook-Group. There are loads of images, both from experienced users and novices. The quality of the images ranges from crappy renders from beginners to top notch from high end guys. Just about the same with every engine, Vray no exception. I have yet to see the real benefit of all this, except for Glare as I explained.

[joconnell]

@tehamnes

There is no such a thing as the Nikon or Canon look: people shoot the cleanest raw they possibly can, and choose cameras based on the inherent quality of their response, these days only differing between makers because of they way each goes about sensors, surely not because of a post tone operation :ie. maybe the primaries are slightly different in gamut shape, that’s all. They ALL, unequivocally, aim at the closest possible RAW capture of light (AKA, in Rendering: LWF), and what stops them from achieving it is physics and technology (ie. it’s a limit, not a choice.).

There’s so a canon and nikon look

All of the sensors are just photon gathering sensors with a coloured fitler over the top for red, green and blue so they’re inherently linear from the off. That of course doesn’t look like film though as film use to have log curves in it’s exposure. The dslr makers mimic this in the post processing of the raw files, they’ve got plenty of metadata in there first of all to take out any inconsistencies in each individual sensor from what should be correct and then second some colour science from the manufacturer to turn the flat linear data into what we’d consider a pleasant photo - not unlike what the tone curves and contrast of fstorm is doing to it’s pixels. There’s been wars started on forums about whether nikons or canons render nicer skin tones for example and likewise the red camera was almost unuseable in it’s earlier generations as with every update of the camera firmware, the colour that came from the camera was completely different and required the DOP to relearn it and shoot brand new wedge tests to see how everything responded. Same linear photon collecting bayer sensors in the camera but a world of difference in what the developing software does to it when presenting a final image to the user!

You’re totally right on the subjective nature of “I think this looks better” too - there’s no accounting for taste and the final needs of the user - if they think the final image in the frame buffer is “pretty” (a subjective view) then hey, great for them, keep doing that - as for you as part of a software development team trying to ascertain what, if anything, is lacking then subjective opinions aren’t exactly the easiest things to act upon

[AlexP]

@3LP

I’d say a huge percent of people pass on images to another point to be worked on

Well I guess it depends on the industry, depends on the studio. I nor anyone I’ve worked with has ever handed over raw renders to someone to work on, but then I dont work in vfx but arch vis. TBH in my time, ive only ever worked with a couple of people that spend more than 10 mins doing post work, just isn’t the time available. Projects only last 4-5 days, a week if you’re lucky.
It has to look good straight out of Vray, so, working lens fx etc are super important to me and everyone else I work with. I only spend about 30secs in lightroom on each image.