Fund to accelerate Development

Discussion related to the LuxCore functionality, implementations and API.
User avatar
lacilaci
Donor
Donor
Posts: 1969
Joined: Fri May 04, 2018 5:16 am

Re: Fund to accelerate Development

Post by lacilaci »

FarbigeWelt wrote: Wed Nov 28, 2018 10:52 am I would probably think differently if I had to use CGI as a professional but then customers would pay for what they ask, namely time, time and sometimes quality. Lot of advertisings based on CGI look average or less compared to what is possible with LuxCore. It just a personal opinion.
This, my friend is what I'm always talking about. It is a bit sad truth but I too face it every day.
"Make it look good and have it done fast!" is coming from every client everywhere, and they do not care about your fancy ultimate and superior algorithms, in the end it simply has to look convincing enough for the client, for me, this means a flexible and easy to use renderer always wins the race and fair comparisons don't matter!

And now, if luxcore dev's do it out of passion and don't care what "market" wants and have their own vision it is all good. But results from "the market" will reflect that.

Here's what my simple mind thinks need to happen with luxcore before it can even properly start to compete with "popular" products (Popular: not better or worse or free or expensive, but popular choice)

1.Clean the mess up. No place for features and modes that break feature parity and only serve very few spectrum of uses.
It's a sacrifice but I do think eventually this will inevitably happen...
get rid of:
bidir - a slow "nope" on cpu which if user enables by mistake and then sees what's going on, next thing he looks for is uninstall button.(of course I'm exagerating this a bit :D)
metropolis - works only good on cpu and really worth it only in the very isolated cases also combined with bidir.. omg just no...
random sampler - only useful for another useless feature(bcd)
bcd - needs replacement, you're wasting your time with this (I'm gonna visit a witchdoctor and sacrifice some animals to make you drop this)
tiled - should be gone or a simple tickbox with no additional settings for cases when running out of vram(but ideally focus on other mem optimization)
light strategy - log power only and a checkbox for DLCS(which should be disabled/hidden until bugs are solved)
clamp - should affect only secondary bounces I can't imagine reason to clamp direct light(but of course this might need some testing)

..Ok so what are we left with now is pathtracing doable on cpu+gpu or either, no setup needed. New user comes hits render and watch the magic happen, professionals can get right to work
(nobody cares about fiddling with settings unless it's a simple slider or checkbox that magicaly solves your optimization problem or unless it's a per object property that allows both artistic and optimization purposes)

2. Fixing shading issues and discprepancies between cpu and gpu needs to take precedence.
One thing is having fakes and workarounds to boost performance when time is precious, but nobody wants to do workarounds cause shader parameter or a bug breaks how rendering result looks, that's a big no no.

3. By this point(after step 1 and 2 is done) luxcore is usable for many many cases even by idiots, also easy to use not prone to user error and not requiring people to do acrobatics with their scene setup.

So we can talk new features now. How?
Clear roadmap that not only shows what is planned but in what order... (I believe it helped corona a lot to have always a clear statement and list (trello - although usually very short list) for every release, because userbase always knows what is the direction and keeps coming back for daily builds)

As for money, I couldn't even navigate through how bounties work. I find it hard to believe it is an effective way of doing funding. Make it easy to fund you through your own website with a big donate button and give a result of how many ppl donated or donate per month (maybe even give an optimal monthly target you want to achieve) and drop every other option so that it is again plain and simple(unless it is actually working for you now, which I couldn't even find out because I'm too stupid)

And before you call me ungrateful and always complaining. I spent over 30min. of my valuable time writing this "list" up not cause I'm ungrateful or decided to piss someone off. But because there is a lot of good things about this project, it is handled by proffesionals, but it is not handled as a product for mass market, so it might always face the same issues over and over again (1.users that don't know how to use it, 2.users that can't rely on it, 3.users that don't find it proffesional due to inconsistencies and lack of flexibility)

Who knows, do you? I might be wrong...
User avatar
Dade
Developer
Developer
Posts: 5672
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: Fund to accelerate Development

Post by Dade »

lacilaci wrote: Wed Nov 28, 2018 2:10 pm Make it easy to fund you through your own website with a big donate button and give a result of how many ppl donated or donate per month
It is all available here: https://salt.bountysource.com/teams/luxcorerender (click on supporters tab)
Support LuxCoreRender project with salts and bounties
User avatar
FarbigeWelt
Donor
Donor
Posts: 1046
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Re: Fund to accelerate Development

Post by FarbigeWelt »

In my opinion, it is not the first place to look in About, to look for Donate.
I think Project is a more appropriate header than About.
What is your opinion?

I agree with lacilaci about ease of use in some parts.
The less options a first user can choose the less curiosity he can have to change for worse. Isn’t it the main approach of common games? Start easy with a few features and slowly increase difficulty with growing skills.

Advanced settings of luxcorerender already point in the right direction. What if these settings have to be enabled in addons user preferences and are hidden by default?
Light and Word designing Creator - www.farbigewelt.ch - aka quantenkristall || #luxcorerender
MacBook Air with M1
provisory
Posts: 235
Joined: Wed Aug 01, 2018 4:26 pm

Re: Fund to accelerate Development

Post by provisory »

lacilaci wrote: Wed Nov 28, 2018 2:10 pm get rid of:
bidir
In contrast I think bidir is one of the most important distinguishing feature of Lux from Cycles.
When you have some time, please do some comparisons, it can be a huge difference, and not at all only for caustics.

Image
User avatar
lacilaci
Donor
Donor
Posts: 1969
Joined: Fri May 04, 2018 5:16 am

Re: Fund to accelerate Development

Post by lacilaci »

provisory wrote: Sat Dec 01, 2018 9:39 pm
lacilaci wrote: Wed Nov 28, 2018 2:10 pm get rid of:
bidir
In contrast I think bidir is one of the most important distinguishing feature of Lux from Cycles.
When you have some time, please do some comparisons, it can be a huge difference, and not at all only for caustics.

Image
First, I don't think the difference between bidir and normal path is only caustics. But 99% of user cases don't really need anything more from bidir than caustics and the speed penalty and feature disparity and not being possible to use it on gpu aswell make bidir not worth using.

Second, what you shown as a bidir vs path comparison is looking more like lack of bounces in pathtracing rather than some dark magic of bidir. Which having gi caching could be eliminated and with great speed as well.
I've worked a lot with maxwell back in the days, and it simply almost never was worth it...

If getting rid of it is too much, then at least put it in some special development/experimental setting so that newcomers don't just expect to switch on a feature and everything works.

Most people/users don't come on forums and argue about these things. The one person I'd recommended this renderer to(at least to try because he likes to try new things) told me on another day that it's a mess and nothing works. And for most people that's the end of the debate, there is no arguing and educating going to happen. This guy needs to work not to learn how and what in renderer works and under which star alignment features work their best.

And this reminds me of the denoiser as well. New guy somewhere on forums posts about denoiser not working too well, then let's say Dade educates them about settings and then the new guy never comes back.... Because he's nice. I've seen this few times, I've seen new people here and there asking around blenderartists, and I decided to not join the discussion to see where it goes and people usually say thanks and we can only imagine if they ever go back to using luxcore...

It's not like I'm against the technology and bidir/bidir+metropolis in general I just don't think it is worth using in most cases. But in luxcore it is not simply switching a button from normal path to bidir and everything works. You really need to reconfigure settings and be mindful of what features you need. And with handful of devs working on this you'll need ages to make every new feature work with every render method in luxcore and yet I'm seeing in feature poll even another one being proposed, this is very inefficient and unrealistic way of handling things imho and going this way the renderer will never be feature complete because not enough devs will join in on this - because there aren't too much professionals interested in the product - because there aren't too much basic users demonstrating strenghts of the renderer- cause it is not streamlined and idiotproof enough - cause feature disparity and lack of direction and consistency...

I might be wrong as always, and I always assume, since devs are much smarter people than me, that they know what they're doing. Maybe I'm just thinking about this too much like it was a commercial product, who knows..
User avatar
B.Y.O.B.
Developer
Developer
Posts: 4146
Joined: Mon Dec 04, 2017 10:08 pm
Location: Germany
Contact:

Re: Fund to accelerate Development

Post by B.Y.O.B. »

lacilaci wrote: Sun Dec 02, 2018 7:27 am Most people/users don't come on forums and argue about these things. The one person I'd recommended this renderer to(at least to try because he likes to try new things) told me on another day that it's a mess and nothing works. And for most people that's the end of the debate, there is no arguing and educating going to happen. This guy needs to work not to learn how and what in renderer works and under which star alignment features work their best.
Yes, one big problem is that we don't have any user testing (controlled test lab, sit 10 new users in front of the program, give them a task to accomplish, record everything and afterwards ask them questions what they found difficult/annoying).
And it is hard to replace this with bug reports and forum threads, because mostly big bugs are reported there and not small annoyances. It seems that people are hesitant to open a bug report just because of the naming of one button or so, when they should do it.

So, a lot of the time we developers can only guess what is best for the users. I sometimes try to switch my perspective by becoming a user again, which has helped me to solve a lot of minor annoyances in the past. I should probably do this more.
But one big problem is that I can never become a new user again. My understanding of light, colors, material properties and render algorithms has changed a lot since I began experimenting with 3D programs. What seems logical to me (e.g. that the sun is a REALLY bright light source) might seem weird to a newcomer ("why do I not see my 17 lumen meshlight in bright sunlight?").

As I said above, almost nobody tells us exactly what they find annoying.
It would really help if you could compile a list of such things and describe how you would change them.
By the way, Sharlybg has done something similar in the past, but I did not yet find the time to test his ideas. So it might take a while.
But if you find a willing developer, it's always possible to start a "user-friendly BlendLuxCore" branch, give it to some testers and see how they like it.
User avatar
lacilaci
Donor
Donor
Posts: 1969
Joined: Fri May 04, 2018 5:16 am

Re: Fund to accelerate Development

Post by lacilaci »

B.Y.O.B. wrote: Sun Dec 02, 2018 8:49 am
lacilaci wrote: Sun Dec 02, 2018 7:27 am Most people/users don't come on forums and argue about these things. The one person I'd recommended this renderer to(at least to try because he likes to try new things) told me on another day that it's a mess and nothing works. And for most people that's the end of the debate, there is no arguing and educating going to happen. This guy needs to work not to learn how and what in renderer works and under which star alignment features work their best.
Yes, one big problem is that we don't have any user testing (controlled test lab, sit 10 new users in front of the program, give them a task to accomplish, record everything and afterwards ask them questions what they found difficult/annoying).
And it is hard to replace this with bug reports and forum threads, because mostly big bugs are reported there and not small annoyances. It seems that people are hesitant to open a bug report just because of the naming of one button or so, when they should do it.

So, a lot of the time we developers can only guess what is best for the users. I sometimes try to switch my perspective by becoming a user again, which has helped me to solve a lot of minor annoyances in the past. I should probably do this more.
But one big problem is that I can never become a new user again. My understanding of light, colors, material properties and render algorithms has changed a lot since I began experimenting with 3D programs. What seems logical to me (e.g. that the sun is a REALLY bright light source) might seem weird to a newcomer ("why do I not see my 17 lumen meshlight in bright sunlight?").

As I said above, almost nobody tells us exactly what they find annoying.
It would really help if you could compile a list of such things and describe how you would change them.
By the way, Sharlybg has done something similar in the past, but I did not yet find the time to test his ideas. So it might take a while.
But if you find a willing developer, it's always possible to start a "user-friendly BlendLuxCore" branch, give it to some testers and see how they like it.
Ok, so I suggest then. Let's first wait for properly working 2.8 blendluxcore as it is now in 2.79. And then I'll put together some design and ideas on how to streamline the workflow and make the tool foolproof.
A separate thread just about usability issues and diagnosing and avoiding user errors... In the end, it might be just about hiding some stuff away from users behind a "expert/devel/test/experimental" setting so that nobody judges the whole renderer because of some inconsistensies.
But changes should be drastic, with corona you just need the render button 90% of the time and new versions of Vray has defaults that too rarely need changing anything. It is also similar story with cycles although I think luxcore could have even more efficient defaults and workflow.

Because in the end pathOcl+sobol gives most feature complete experience(relatively) and luxcore in that config is pretty much easily usable for many things with great performance. Not to discourage from using and experimenting other methods, but their presence in UI should reflect their usefulness and reliability. 90% of users should not even need to touch render settings and be made aware that defaults are the best.

It is also how some things worked in early corona days. Where there were a lot of configs in "devel/debug" category (including bi-dir, vcm, sppm and some weird things that weren't working the best)
User avatar
lacilaci
Donor
Donor
Posts: 1969
Joined: Fri May 04, 2018 5:16 am

Re: Fund to accelerate Development

Post by lacilaci »

I've created separate thread for workflow related issues and ideas, rather than turn funding ideas into that: viewtopic.php?f=5&t=774
User avatar
Dez!
Posts: 368
Joined: Sun Apr 08, 2018 1:09 am
Location: Ekaterinburg
Contact:

Re: Fund to accelerate Development

Post by Dez! »

My monthly income is 530 dollars.
I can contribute 20. And I will do it with joy.
Linux Plasma | Ryzen 5, 32Gb, SSD M2, GT 590 RX | BenQ 27 | Wacom One | Microsoft Ergo | Tie Guan Yin tea
http://dezigner.tilda.ws/
Post Reply