Login - Create Account

The Next Big Thing

Login or Register to post new topics or replies
Vladimir Golovin
Administrator
Posts: 3316
Filters: 55
There are many great ideas that circulate in the Filter Forge community and in the wider world about procedural content generation and usage. Many of these ideas are great, but implementing them into Filter Forge is problematic.

A good example of such an idea is GPU-based rendering: it offers obvious benefits and everyone wants it, but there are numerous technical obstacles to porting FF on GPUs in its current form, the primary one being the rasterization-based (as opposed to raytracer-based) architecture of mainstream GPUs.

Another such idea is Bezier splines. It’s the founding feature of the majority of graphic design apps today, and it lends itself to procedural paradigm pretty well, but Filter Forge, with its narrowly-specialized nodal language can’t currently handle operations with integers, floats and lists that are needed for constructing splines (or, to put it simply, it lacks Turing-completeness).

The question: How to make use of all these ideas without breaking Filter Forge?

My answer: do something new from scratch, as modern as possible, applying all our experience with Filter Forge.

The question is, what is this next big thing?

We’ve been working on finding the answer for several years. We now have a five-person team for prototyping and R&D: a language developer, a GPU programmer, an iOS programmer, a professional interface designer and a mathematician.

Below is some assorted info on our progress so far:
  Details E-Mail
Vladimir Golovin
Administrator
Posts: 3316
Filters: 55
What are we working on?

The general idea is to create something that allows anyone to design good designs (business cards, websites, logos, ads, signs, banners etc.) without being a design professional. The system will be centered around 2D (non-3D) static (non-animated) grid-based graphic design.

The core paradigm is high-level graphic design. In the current, low-level graphic design, the designer deals with rectangles, circles, splines, etc. In our new paradigm, the designer deals with cats, arrows, houses, skies, cars, flowers etc -- each one of these is a parametric object with adjustable colors and other parameters. A rectangle, a triangle or a circle is just a special case of a parametric object.

Technically, it will be implemented as a symbiotic system of two apps (one for content consumption and another one for content authoring) and an asset store for buying / selling content.

The consumption app is for arranging pre-made smart parametric objects into designs. It will be fully cross-platform, including mobile OSes such as iOS and Android, and Windows 10 via their universal binary for all form-factors, and will support both mouse-based and touchscreen interfaces.

The authoring app is for creating smart parametric objects. It will be desktop only (but I’m afraid we’ll have to support touch controls as well, because Windows 10), and will resemble a hybrid between a vector editor and a node-based programming environment. Think Houdini for 2D, or Filter Forge for vectors and much more.

Here's a preliminary design for consumer app for iOS:

  Details E-Mail
Vladimir Golovin
Administrator
Posts: 3316
Filters: 55
Technology

Here’s the technology outline for the entire system.

Implemented:

- The general approach is vectors (mostly cubic Bezier splines), not pixels.
- Shapes are generated procedurally using a node-based visual language.
- Shapes can have arbitrary GPU-supported fills (see Shadertoy for amazing examples of fills).
- GPU fills are designed nodally / visually, think mini-Filter-Forge.
- Blending between shapes is severely limited. Blend modes yes, refraction / distortion no.
- HDR colors are supported (with some limitations).
- Support for multiple external bitmaps with resolution up to 8000x8000 pixels.
- Bicubic bitmap filtration is supported both in the preview and the export.
- Tile-based rendering and export, so output resolution should not be a problem.
- Can render to floating-point HDR formats, including OpenEXR.
- Target hardware must support OpenGL ES 3.0.

Not yet implemented / undecided:

Bitmap-based effects (e.g. Gaussian Blur, Median) are not implemented yet. I currently see no clear pathway for properly implementing them in a system I outlined above. Is this a showstopper to you? Would you use this system if it lacked blurs?
  Details E-Mail
Vladimir Golovin
Administrator
Posts: 3316
Filters: 55
Our approach to GPUs

Basically, there are two approaches to GPUs.

The first one is boils down to raytracing. It gives us absolute power in image generation. It generate anything: force fields, rainbows, 3D, refractions, whatever. The downside: it’s slow, not all devices support it, doesn’t natively fit the GPU architecture (there are raytracing accelerators such as that PoverVR chip, but they are far from being mainstream yet).

The second approach is native / naive: just make the GPU do what it does best, which is drawing filled polygons on the screen. Upsides: extremely fast, easy to implement, low hardware requirements (can be implemented on today’s mobile devices). Downsides: potential output inconsistency across devices (due to differences in fixed function hardware parts of the pipeline such as anti-aliasing, blending and texture filtration), blending between polygons is severely limited, distortion / refraction effects that are effortless in a raytracer are hard to implement and prone to artifacts.

We decided to go with the second, native / naive approach. One of the reasons is that we want our tool to work on as much computers as possible. Raytracing requires good hardware, and there’s no way to implement a CPU fallback for users who have weak hardware. Naive, on the other hand, doesn’t need the CPU fallback codepath because GPUs that can handle it are everywhere today. Our target capability level is OpenGL 3.0, which is already pretty common these days.

(If raytracing-capable hardware ever becomes mainstream, we can switch our renderer from rasterization approach to raytracing approach, in a year of work or so.)
  Details E-Mail
Vladimir Golovin
Administrator
Posts: 3316
Filters: 55
New node-based language

The picture below shows our experimental visual node-based language.

The language is Turing-compete, which means you can evaluate anything in it. It supports any number of types (as opposed to Filter Forge's Green, Blue and Gray). There’s a limitation though: it’s a pure functional language where programs are evaluated immediately and momentarily (no concept of waiting and time), so you can’t implement a game in it. Another limitation: you won’t be able to use it inside GPU fills / shaders. The shaders will be created via a separate visual sub-language that is similar to Filter Forge.

The example below shows a program that takes a list of integers from 1 to 10, flips the sign of each its element, then adds 7 to each element, then takes 3 top elements of the resulting list and outputs that as a result.

(The code snippets are just for reference / comparison, they are not part of our nodal language. The topmost one-liner example is Haskell, the three-liner below is a Lisp, namely Racket, and the node tree is our visual equivalent.)

  Details E-Mail
Vladimir Golovin
Administrator
Posts: 3316
Filters: 55
And here’s another simple example:

  Details E-Mail
Vladimir Golovin
Administrator
Posts: 3316
Filters: 55
Here's an example of an explicit point-by-point spline creation for the discussion with Crapadilla in a neighboring thread. In this example we evaluate a list of points for constructing a star:

  Details E-Mail
Vladimir Golovin
Administrator
Posts: 3316
Filters: 55
Why am I posting this?

I was working on this thing full steam ahead, and then I hit the wall. The wall is psychological. Basically, I started asking myself all these startuppy questions, the most important ones being “What user pain does it solve?” and “Who will use it?”

Do you find the system I outlined above appealing to you? Or, more precisely, would it deliver real value to you beyond its obvious “toy value”? Or is it a solution in search of a problem?

How would you shape it to suit your needs / wants more?

Does the new nodal language look too scary to you?

And, finally, would you use such system if it lacked blurs? smile:D
  Details E-Mail
Vladimir Golovin
Administrator
Posts: 3316
Filters: 55
(Hey Sign Guy, I see you reading this thread. I'd really appreciate your honest, no-holds-barred, no-punches-pulled opinion on all this.)
  Details E-Mail
ThreeDee
Lost in Space

Posts: 1666
Filters: 111
Hi Vladimir,

Personally I wouldn't have use for a high-level program for placing ready-made images, and I have hard time understanding how a node-based method would make it easier than a simple visually oriented drag-and-drop system like in most layout programs.

However, a node based, procedural/programmatic way to create vector art (a "Filter Forge for Vectors") would be interesting and likely of use to me.

And what would be of extreme interest would be a node-based and/or procedural/programmatic method of creating 3D models. I would pay big bucks for such a program.
  Details E-Mail
Sign Guy
Digital Art Developer-Publisher

Posts: 418
Quote
Vladimir Golovin wrote:
(Hey Sign Guy, I see you reading this thread. I'd really appreciate your honest, no-holds-barred, no-punches-pulled opinion on all this.)


Hey Vlad,

I lightly read your posts (as opposed to skimmed) and appreciate the parts where you took off your programmer's hat and tried to put into terms of what uses and issues I could understand. My eyes tend to glaze over and I lightly snore when it gets into programming jargon. I'm a graphics guy and not a programmer. All I want to know about is how you will save me time or allow me to do something I can't do now that will put jingle in my pockets.

That said, I will add that I haven't been using FF much due to its slow rendering and leanings towards procedural creation rather than photographic image manipulation.

As to splines (vectors are the more commonly used terminology for sign people), there are two things that make them distinctive and useful. 1) They are resolution independent and can be resized with no effect on quality. 2) They can be used to control the X, Y and Z axis movement of machines from plotters, to CNC routers and machining devices, to 3D printers. Other benefits are far less important.

Anything aimed at making layout and design easier for novices and non-professional users is, IMHO, pretty well covered and will tend to get you ignored in the various professional communities. The same is true for special effects that get over used and abused be people lacking in experience, training, supervision and talent. Many printing and sign making professionals view this sort of thing as a detriment to our industries. The other side of the coin is that products aimed at professional users are highly sought after and bring substantially greater prices. The typical sign business, for example, spends more every month for pricing and order management software subscriptions than FF charges for an owned license.

So the first question is: What market do you want to develop software for?

If for the professional market, then I would recommend joining and exploring a few forums to get an in depth understanding of what they might need. I invite you to join and explore my forum, Signs101.com. We are the highest traffic sign industry website on the internet and have a demographic of at least 95% of our users being professionals. Once you develop a feel and an understanding of our members, I would be happy to introduce you and have you question our members about anything you have in mind.

You might also benefit from exploring CNCzone.com who is a leading forum for machinists and CNC production.

Questions? Fire away.
Fred Weiss
Allied Computer Graphics, Inc.
  Details E-Mail
ThreeDee
Lost in Space

Posts: 1666
Filters: 111
Filter Forge for Vectors

I’ll take a new look at this from the viewpoint of creating a fully vector-oriented “Filter Forge for Vectors”.

Input, processing and output would all be vector, and all processing would be procedural. And it would be with low-level control.

While you could consider the options that all vector information could be created from scratch in “FFV”, it would be a limitation I would hesitate to impose. It is like leaving the External Image option out of Filter Forge, and would cut out half the functionality.

Input:

-You would have to be able to import (a number of formats of) vector art. Or actually multiple files.

Creation:

-You would have to be able to create basic geometric shapes and freeform bezier splines.
An important addition (at least for me) would be the ability to generate the bezier spline programmatically (which could be node-based or script).

Processing:

- You would have processing steps for entire shapes like transform, boolean, repeat, offset path, warp, blend, etc.
- Recursive options would be nice…
- You would have processing steps for individual vertices, like add vertices, select (with specific conditions, such as "every third"), transform, randomize position, scale (with certain origin point), etc.
- You would have stylize options (drop shadows, fake 3D, bevel)
- I wouldn’t care for bitmap processing, since the target would be vector output.

Drawing:

- You would have drawing components that set the fill and stroke values for objects, as well as blending modes.
Here I’d like to see variable line weight options. Different fill and draw styles might be interesting (fill patterns, brush strokes) but they also increase vector information massively.

Output:

- You would have a Result component to show the output.
- You would have to be able to export as (a number of formats of) vector file. (Option to export as bitmap.)
- Tiling: both seamless tiling and non-seamless tiling options. The output would generate as many tiles as you set it at (like Size, Tiles style)
- Should be able to set a specific physical output size.

Probably not much new there you haven't already thought of.
  Details E-Mail
Rachel Duim
So Called Tortured Artist

Posts: 1173
Filters: 73
I would use it. "Scary" to some users? Probably. I like the compromise between regular FF node programing and the map/curve scripts in LUA. Although not immediately obvious, this will provide more "direct" access to variables as opposed to doing all the math in RGB or complicated scripts. And that can be very powerful. Blurs? Ha. I think that you will get some non-programmers to use it, and programmers like myself will like the convenience and the speed of the node based approach. Am I making any sense?
Math meets art meets psychedelia.
  Details E-Mail
Vladimir Golovin
Administrator
Posts: 3316
Filters: 55
Quote
And what would be of extreme interest would be a node-based and/or procedural/programmatic method of creating 3D models. I would pay big bucks for such a program.


ThreeDee, have you tried Houdini?
http://www.sidefx.com/
  Details E-Mail
Skybase
2D/3D Generalist

Posts: 3907
Filters: 75
Hello!

Thought I'd take a bit before I drop off my thoughts on it. I first off very much welcome the idea and the work that's being done. It sounds spectacular and knowing that there aren't many products like this, I think it's going in a interesting direction.

While there's little that I can specifically comment on, I can first off say that the project reminds me of Nodebox 3 (https://www.nodebox.net/node/). Just the nodey aspect of it and the fact that it handles vectors etc. Pretty sure the product direction is completely different.

As for some of the questions listed:
1. I would use it only if it serves a purpose to my workflow. I work in the motion graphics industry and we constantly need assets. It's the reason why FilterForge is so useful. This would also lend itself to some of the game dev I'm doing. This kind of product is probably what we're all interested in as a designer.

2. It's not a toy as long as you include the necessary "professional tools", aka what the industry generally sees as being professional. If it's vector-based authoring, those "tools" would be stuff like alignment, the ability to fix minor issues, ability to export to specific formats or templates etc. The point here is to avoid the tool becoming a cut-n-dash utility where people just slap down what they want and move on. Basically, the tool should obviously let people do what they want, but also get to the details somehow. I personally get extremely picky with formatting, color, line weight, spacing (distances between objects) and so on, but this is something we need as designers. Precision is key, but also the ability to flexibly produce unexpected results is also a welcoming thought.

SignGuy totally nailed the thought with CNCs. I'd say the DIY stuff is lately so powerful that it's kinda hard ignoring it. I would probably use tools like what's described above this to accomplish complex tasks... I mean the description doesn't sound like it but there's the potential for it.

3. The node network doesn't look scary to me, but it's probably because I've become very used to seeing big complicated scary things.

For that kind of thing I would say. 1. Have documentation that outlines details (duh) 2. Make professional-grade examples with annotation. 3. Have beginner-friendly tutorials that cover basics... then advanced stuff as well.

I feel a lot of people on FilterForge get stuck somewhere with the technicalities. We're all on different pages with our lives and we all got different ideas. I personally think sometimes the weakness of FilterForge is simply that there's just a variety of people contributing, leading to a huge spectrum of excellent to poorly developed filters. I have no exact idea of what the end product here will be so I'm basing this off of what's written above.

Overall, I think being resource rich would help with the "scary" factor.
  Details E-Mail
Vladimir Golovin
Administrator
Posts: 3316
Filters: 55
Quote
3. The node network doesn't look scary to me, but it's probably because I've become very used to seeing big complicated scary things.


Skybase, the good thing about the examples I posted is that they show the nodal language at its scariest. I mean, usually you'll be dealing with something much simpler, on much higher level, like you do in Filter Forge -- but if you want a custom computation, nothing prevents you from going all the way down to language primitives such as Map / Filter / Fold and data structures, which is precisely what these examples show.
  Details E-Mail
Vladimir Golovin
Administrator
Posts: 3316
Filters: 55
(Fred, thank you for your reply. Sorry for my long silence: you asked a lot of great questions and I'm now meditating on them in a Google Doc with a copy of your post).
  Details E-Mail
Vladimir Golovin
Administrator
Posts: 3316
Filters: 55
Quote
Blurs? Ha.


Rick, could you please clarify? smile:D You mean that not having blurs is not a big deal?
  Details E-Mail
Rachel Duim
So Called Tortured Artist

Posts: 1173
Filters: 73
The more I use FF the more I see how blurs slow everything down. Ban all bitmaps. Call it smoothing instead smile;) A necessary evil, eh? Even with curved vectors...
Math meets art meets psychedelia.
  Details E-Mail
FFCreator
FFCreator
Posts: 36
Quote
In our new paradigm, the designer deals with cats, arrows, houses, skies, cars, flowers etc
..... and an asset store for buying / selling content.

Reminds me of CorelDraw (vector based drawing program). They sold a 100.000+ library with vector objects (on CDs). Nowadays Inkscape (free) can handle vector objects.
I don't think I would spend money for regular content or another regular composer. Why I am using Filter Forge? Because you can experiment in depth and make unique things, because of the huge free database with content; not to compose high level vector objects bought from an online store. But maybe I am not the target audience (hobbyist) or I don't understand your plan.

The rules of software development seem to be that at a certain moment developers want to start from scratch. This is not easy and most of the times it takes years to fullfill the needs of their existing users. See Sony Vegas / Catalyst and Adobe with their subscription model.

My need for speed with Filter Forge is when I am experimenting with FF. If the final rendering takes a little longer, it is less distracting. What about only adding GPU accelerated equivalents of the existing FF components or adding an extra mode "quick and dirty"?
Advantages : backwards compatible, no start from scratch and for the existing users quick rendering in "quick and dirty" mode.
  Details E-Mail
ThreeDee
Lost in Space

Posts: 1666
Filters: 111
Quote
Vladimir Golovin wrote:
ThreeDee, have you tried Houdini? http://www.sidefx.com/


I haven't, but from the looks of it I may well do so soon. Thanks for the tip!
  Details E-Mail
StevieJ
Designer/Artist

Posts: 11262
Filters: 163
Hi Vlad... I just read through...and I would agree that there is too much out there already...and a vector FF would make much more sense to pursue...especially towards creating programs that interconnect with FF... I think vector-based is a good place to start...

Further, I would suggest expounding upon the 3D realm in vector modeling that is simpler and more cost effective to use than the other programs out there...which you are hedging upon in your initial idea...and could use third party rendering programs like Vray...

Finally, I would suggest the development of a 2D and 3D algorythm creation and rendering program that picks up where Apophysis left off...or even use the open source Apophysis program itself to build upon it... As a fine artist, I still use Apophysis, but I wish there was a program that expanded its true potential while retaining its ease of use... I mentioned this years ago because it seems like it goes hand in hand with FF...and would make sense in a suit of programs with FF... Just my off the cuff thoughts about it... smile:)
Steve

"Buzzards gotta eat...same as worms..." - Clint :)
  Details E-Mail
Sign Guy
Digital Art Developer-Publisher

Posts: 418
I'm not into 3D but I am very active in making seamless tiles ... especially from photographic materials. Towards that end, I think Genetica is head and shoulders more capable than Filter Forge. For whatever the reason however, Genetica has all but been abandoned by its developer (although still available for licensing at their site).

Here's a few examples of photographic seamless tiles I've created:

Basketballs

Jelly Beans

Butterflies

Three different methods are provided for tile creation: Synthesis, Patches and Patterns. The Patch method was used for each of the examples shown.

My suggestion to you is to have a close look at the capabilities of Genetica which also include vectors, animation, and more. Then contact the developer and talk about a merge/takeover or a buyout.
Fred Weiss
Allied Computer Graphics, Inc.
  Details E-Mail
Vladimir Golovin
Administrator
Posts: 3316
Filters: 55
Fred, regarding seamless texture synthesis from bitmaps: the problem with integrating these algorithms into the current Filter Forge stems from the fact that it doesn't properly support finite bitmaps. Everything in FF is represented as an infinite / endless images, even regular bitmaps. This is due to a stupid architectural decision I made almost 10 years ago. If FF supported finite bitmaps, implementing these algorithms would be trivial.

We'll definitely going to rectify this in the FF successor, which will include first-class support limited bitmaps with known dimensions.

(I probably should post an update on my revised approach to The Next Big Thingy: In short, I dropped all mobile ambitions and narrowed the approach to a powerful, Turing-complete, GPU-assisted Filter Forge successor; but that's a story for a separate post).
  Details E-Mail
Skybase
2D/3D Generalist

Posts: 3907
Filters: 75
Bump! Curious to know what's happened since last year!! smile:D
  Details E-Mail
SpaceRay
SpaceRay

Posts: 9762
Filters: 33
Quote
Skybase wrote:
Bump! Curious to know what's happened since last year!!


Me too, because this looks to be very interesting and great, what I do not have it clear, is if this new thing will replace filter forge, or it will both different things existing and evolvimg together

The problem I see is that I suposse and think that none of the filters of filter forge will work on this new thing as if this is totally new and rebuilt from scratch it will not be compatible, although I wish I was wrong
  Details E-Mail
Vladimir Golovin
Administrator
Posts: 3316
Filters: 55
Just a quick upd ate on what's happening:

1. We're still deep in R&D for the Turing-complete node-based language. Now working on mutually-recursive functions and polymorphism. All this is scary stuff, especially considering that I'll be the one to turn it into a product.

2. I think I decided on the overall GPU approach. We have separate subsets of nodes / components / functions that will serves as non-Turing-complete "domain specific sub-languages" for handling the GPU. For example, you could have a set of components for CUDA and a separate se t of components for Vulkan. The GPU components won't be Turing-complete, but they will be embedded in a Turing-complete "host filter / program".

All this sounds a bit complex, but all this complexity is on the side of filter authors, not users. And I think this is OK. Good filter authors hit the 'possibility ceiling' of the program pretty quick, so I think offering much better expressivity by taking away some simplicity is a good long-term approach.
  Details E-Mail
SpaceRay
SpaceRay

Posts: 9762
Filters: 33
Thanks very much Vladimir for answering and giving more details

Love much the idea that it will be using GPU acceleration as this changes everything and really makes it very usabke and fast, great!

I know that you may not know it, by may you have a rough idea of when this could be released, finished and available?
I mean maybe end of this year, sometime in 2018, or for 2019?

Is there a plan to make a beta testing?

I wish you and all the team behind this all the best and that you can make a great work

Thanks
  Details E-Mail
Skybase
2D/3D Generalist

Posts: 3907
Filters: 75
While words may have flown over people's heads the idea is exciting to see how you're constructing and developing a complex system. This is the sort of thing that normal people won't see and it's just something so core yet so hidden from the world at large. It makes me appreciate the time and dedication that it takes to produce one product and not only that, it's insightful and unique.

Hope to hear more!
  Details E-Mail
Vladimir Golovin
Administrator
Posts: 3316
Filters: 55
SpaceRay,

Yes, GPU support is a necessity, an absolute must these days. One of the problems with the GPU is the API choice under macOS. Apple is notoriously hesitant when it comes to implementing modern GPU APIs, such as Vulkan, and on the CUDA side thins aren't perfect as well. What I know for sure is that I definitely wouldn't want to develop a separate set of components for Apple Metal.

Regarding the timeline or release date - honestly, no idea. I wanted to finalize the node-based language by the end of 2016, but the guys (and girls!) are still stuck deep in polymorphism, type inference and mutual recursion. Plus, after that, we'll have to productize the entire thing, which is far from easy. So, it's done "when it's done" ™.

Beta testing - yes, absolutely. I'm even thinking about releasing a 'playground' app without GPU and image-processing functionality just to let people play with the node-based language. What's more, I even considered open-sourcing the language, but it's way too early to think about that.
  Details E-Mail
Vladimir Golovin
Administrator
Posts: 3316
Filters: 55
Skybase - thank you smile:)

Yes, there was a moment when I clearly saw a strong connection connection between Filter Forge and a typed, lazy, pure, functional programming language - namely Haskell. I'm not alone in this, there were at least two attempts by other people to create a visual / node-based syntax for Haskell, but I didn't like their results. And we're not creating a syntax for Haskell - we're creating a visual programming language that is very strongly influenced and inspired by Haskell.

However, all this is scary. There's a Russian saying, "shooting sparrows with a cannon". Sometimes I feel like we're inventing a superconductive railgun, complete with a cold fusion plant to power it, just to shoot at image-processing sparrows.

(On the other hand, if we pull this off, this thing will feed us until the end of our days, because no sane company in their right mind would compete with us smile:D)
  Details E-Mail
kirkl13
Posts: 30
Surprised nobody mention Substance Designer on this forum. Long ago I started with Filter Forge but now Substance Designer is my main tool as a gamedev artist. It fulfilled very same demand but on much more satisfying level, basically because of speed, gpu support and immediate realtime preview.
Still I miss a lot of FF conveniences, like Bomber and super easy and scalable noise nodes, tileable on demand, scale, image size and form factor independent. Not necessarily square textures, automatically adjusted for a new scale. True HDR tricks being really possible.

To be honest I hate Substance Designer. Imo it seems as if it had been done by some alien humanoids, not humans. So much pain in the a.. for things that should be like 2x2. But through years they fixed many of its initial flaws and now it overweights FF so much I can't even imagine to be going back.

So before doing something new and bizarre I would offer to make a true Substance Designer rival, contrary to Substance Designer being really convenient and artist friendly, keeping all the advantages FF has plus real time-visual feedback and node system borrowed from Cycles/ Blender. In fact I would be perfectly happy if FF work as a set of nodes for Blender. Would buy such plugin instantly.

But most of all I need something like "slope blur" node from Substance Designer. Spent hours to figure out how to do it in FF and not taking days to calculate. Still unsuccessfully.
  Details E-Mail
PayPaul
PayPaul
Posts: 71
Quote
kirkl13 wrote:
To be honest I hate Substance Designer. Imo it seems as if it had been done by some alien humanoids, not humans. So much pain in the a.. for things that should be like 2x2. But through years they fixed many of its initial flaws and now it overweights FF so much I can't even imagine to be going back.


From what I've seen of its website, Substance designer is subscription based. I would want to OWN the program, not pay through infinity for it.
FF and the lifetime license I own allows me infinite future upgrades and I look forward to those enhancements and improvements over time. I despise subscription based programs of any kind.
Don't step on a dead dogma. It'll turn up again and bite you!
  Details E-Mail
Vladimir Golovin
Administrator
Posts: 3316
Filters: 55
Quote
kirkl13 wrote:
To be honest I hate Substance Designer. Imo it seems as if it had been done by some alien humanoids, not humans.


I honestly tried to take a closer look at it, several times, but I just can't get past the inteface. It just baffles me, the same way Genetica did.

Aside from the GPU support (yes, I know) and Slope Blend, what things do you miss in Filter Forge, compared to Substance Designer, as a gamedev artist?

(BHTW, we will be introducing some things that make life easier for gamedev artists -- for example, in FF7.0 we will have simultaneous rendering and export of all channels, such as bump, normal, diffuse etc. Also, we rewrote the renderer so in the future it will be able support other shading modes for the Result Component, e.g. Unreal or Unity PBS -- but this won't be out in FF7.0.)
  Details E-Mail
SpaceRay
SpaceRay

Posts: 9762
Filters: 33
Quote
Vladimir Golovin wrote:
Aside from the GPU support (yes, I know)


This is the main reason I do not use FF anymore as before, because is the only software I have that is takes much time just only to PREVIEW the result. I have to say that I really love FF and is really awesome and amazing what you can do, but regretably in 2017 is not justified to use a software that takes so much time to even preview the result ( at 600 x 600) and then even more time when you want a higher resolution

And maybe is the reason why the forum is so empty and inactive, as in 10 years all the other software have made things to accelerate and made its software faster, but FF is only slightly faster and the speed depends as always very much on the filter you use and the output resolution size

SUBSTANCE DESIGNER PAY TO OWN OPTION

Quote
PayPaul wrote:
From what I've seen of its website, Substance designer is subscription based. I would want to OWN the program, not pay through infinity for it.


I am sorry but you are wrong in that Substance Designer is subscription based only if you are an Indie (earn less than 100.000 a year),
because there is an option as that is PAY TO OWN, I mean that in this one they divide the final price into 12 monthly payments, so each month that you pay you are nearer to OWN the software, and when you finish paying it, is optional to keep paying for updates or not, as you get a perpetual license

So it is 12 x 19.99$ + 49$ = 289$ to own the software

Quote
How many payments does it take to get a perpetual license for the software?

This option is available only for Indie licenses.

When you are subscribed for more than 12 months consecutively, when you cancel you get the option to purchase a perpetual license of all the software for $49.

This license does not provide any maintenance to the software. You can purchase maintenance for each software on the website.


If you continue paying each month you get the updates for free and also the maintenance, but you are not forced to do it if you have already a perpetual license
  Details E-Mail
Skybase
2D/3D Generalist

Posts: 3907
Filters: 75
Quote
Aside from the GPU support (yes, I know) and Slope Blend, what things do you miss in Filter Forge, compared to Substance Designer, as a gamedev artist?


I found that the Distance node in Substance is extremely useful for many cases and is in fact an essential part of the workflow.
  Details E-Mail
kirkl13
Posts: 30
Quote
Vladimir Golovin wrote:
Aside fr om the GPU support (yes, I know) and Slope Blend, what things do you miss in Filter Forge, compared to Substance Designer, as a gamedev artist?


At first I would say Substance Designer still rely a lot on CPU I believe. At least I see no dramatic GPU loading fr om it at all and my last video card upgrade gave me nothing . It sometimes slow as hell , same as FF.
But really huge plus is ability to work with real-time material preview. Including kind of geometry tessellation to show surface height as actual displaced geometry. All this with modern PBR shader everybody use in CG now. For uncertain reason not modern DX11 displacement/tessellation one you can preview in Unreal editor but still pretty close.
I bet it's what it does with GPU mostly, not actual node calculations, although not 100% sure.

Still the real time geometry displacement/normal map preview is what makes it super convenient. I stopped to use Zbrush to sculpt anything .

To list what I miss:

1. Most of all an ability to do "Slope Blur", not blend, while in fact it's really blend of gradually fading noise distortions around contrast elements. Also "Bevel" same as Photoshop outer/inner glow. Tried something similar in FF and it made it unusable. Slope Blur is a kind of distorted outer glow.

In FF both things work ok around polygon shapes only, In Designer at anything, any contrast black and white image to simulate kind of depth erosion. And you could use curve editor to control precise geometry shape/fading It's a huge thing I miss in FF wh ere I can do same on limited patterns, polygon shape edges only.

It's most important reason I moved to SD actually.



2. ability to save part of the node graph as my own tools and quickly drop them fr om side panel library.


3. Nodes auto connecting ( Blender nodes do it much better actually)


4. lots of "ready" tools like "height blend" "lighting cancellation", "make it tile" so you even don't have to do your own.


ps. You really should look into Substance Designer. It's CG industry standard now. Nobody use Genetica now and only few still remember what Filter Forge was in modern Game industry.
Even Photoshop now is a kind of secondary tool for photo correction only.

I still love a lot of easy things in FF: Bomber , true HDR mode close to what hardware shaders do, much easily adjustable input noises. In many respect I still feel home in FF and always feel sad it have not evolved enough to be truly useful.

Would love to have vector image editor using a combination of bitmap and procedural fills, being depth or "deep pixel" aware in terms of Nuke . Wh ere objects do realtime on screen Z depth combine plus procedural variations.

A kind of 2d vector image composer to mix procedural things with 3d rendered elements and at the same time have layer based interface wh ere we could easily select something of screen , regenerate/add procedural things like wear and rust and easily move around and scale like in Photoshop.
On front layer being easy and simple to manage manually still allowing to re-construct things with nodes deep enough. Not turning into undecipherable mess and Gordian knot of nodes connected through crazy invisible "exposures" Substance Designer is.

ps. If you are self employed contractor and make less than $100k Substance Designer is cheap as dirt on Steam without any pay to own or subscription. Wouldn't be surprised if they will rise the price since there is no true competitor for them currently.
  Details E-Mail
Vladimir Golovin
Administrator
Posts: 3316
Filters: 55
Quote
4. lots of "ready" tools like "height blend" "lighting cancellation", "make it tile" so you even don't have to do your own.


Could you elaborate a bit on these?
  Details E-Mail
kirkl13
Posts: 30
Quote
Vladimir Golovin wrote:
Could you elaborate a bit on these?


Substance Designer has so called "atomic" base nodes like Distance mentioned by Skybase and library tools nodes , a saved node graphs doing certain image manipulations. Like groups in FF. But they are already there and done properly without mistakes or inefficiencies a user could do on his own.
Nothing I couldn't do in FF usually except that "glow" thing which freeze FF forever and I am not sure, maybe it's just my mistake and I do something wrong trying to reconstruct it in FF .

You can open an inner graph of such tools and figure out how they was done, theoretically. Although in fact the crazy SD node structure makes it often very hard to figure out anything.

https://support.allegorithmic.com/docu...ight+Blend
https://support.allegorithmic.com/docu...requencies
https://support.allegorithmic.com/docu...Tile+Photo
https://support.allegorithmic.com/docu...Slope+Blur

As I said I can do all the same in FF easily except the Slope blur which in fact my most frequently used thing in SD. In FF the "Smudge" is somehow close but so slow it kills everything.

I bet they use just kind of tricky simple workaround probably.

Ironically for years I had begged Allegorithmic to look at Filter Forge and do a few similar nodes and finally they did, like recently added Curve node. So I have less and less incentives to go FF.

Still Substance Designer annoys me non stop with its node system craze, "exposures", "functions" "related to parent" etc. I would happy to see a true rival with human made node system closer to traditional ones, like what Blender has for example. Maybe a kind of commercial Blender plugin.

ps. Sadly I am not even sure if I am going to upgrade to FF 7. Contrary to many artists in gamedev I still use Photoshop a lot and would Filter Forge work as non-destructible and ever editable filters for Smart objects without rasterizing them into regular layers and able to read other layers as bitmap inputs, I would instantly buy it. Even if it would be a bit slow it world revolutionize Photoshop itself and would turn it from being ancient junk to truly modern soft. But I am afraid it's not possible without true GPU support probably.

I would soon ditch Photoshop altogether probably if FF wouldn't work that way and will be using Affinity Photo instead.
Would FF work as "live" effect for Affinity Photo and Designer, it would be an instant buy too. I remember FF worked that way for vector editor Xara once. Not anymore.
  Details E-Mail

Join Our Community!

Filter Forge has a thriving, vibrant, knowledgeable user community. Feel free to join us and have fun!

29,459 Registered Users
+17 new in 7 days!

137,914 Posts
+7 new last day!

13,449 Topics
+8 new in 7 days!

Online Users Last minute:

10 unregistered users.

Recent Wiki Edits:

Follow filterforge on Twitter