Is FF Inc. making something for FF 4.0 to have a FASTER render engine?

I think that Filter Forge is a very good software, and will be excellent when 4.0 is released.

Please, I want to know if the FF Inc. is working on the development of a optimized and better render engine that could process the filters faster than what is available in FF 3.0.

One of the main problems I see is that although I have one of the fastest computers available (Intel Quad Core i7 2600 3.5 Ghz Sandybridge Z68 16 GB RAM) Filter forge keeps being SLOW on lots and many of the filters that you want to use with a higher resolution with at least 3000 x 3000 pixels with really is not very high for printing the results. I think that there is no use for a 600 x 600 resolution.

I will be making when I can some test as a PROOF that FF 3.0 even on a very fast and modern CPU multicore (8 cores= 4 cores + 4 virtual) KEEPS being SLOW.

It is FAST on simple filters or ones that do not involve complex calculations, for example on photo filters effects is fast.

It would be a very bad thing if FF 4.0 would have the same render engine as FF 3.0 with the same speed.

I hope and wish that the FF developers team have some plans to optimize and make a faster render engine on FF 4.0, otherwise FF will sink compared with the graphic software competition that will be available at the time of the release of FF 4.0.

Ok well, as much as I see a point in this, I also think we shouldn't jump to immediate conclusions because there are a number of things you can do to save yourself render time. And while I do agree to a point about how some things in FilterForge are sluggish I find the "sluggish" issue present in many other programs that I use frequently. And with a little bit of background knowledge, experience, and understanding of these things you can potentially reduce render times.

So let me begin some lecture!

At least fr om what I understand, the current progressive render in FF3 is quite fast. It gives you visual feedback relatively quickly compared to previous versions and that makes it easier for you to make decisions so you don't have to sit there all day waiting for that texture to come out right.

Let me mention a couple things that slow your render down. First of all, depending on anti-alias settings, you can really get slow results. Does it need full anti-aliasing? Or can you get away without it? Diagnose AA zones: look for unusual edges or just things that look wonky. You may be able to go into your filter and solve those issues.

Ambient Occlusion and Reflective Occlusion: if you set these to super high numbers, you'll obviously get super slow results. Make sure you test your AO settings before rendering a full image. Don't assume "high numbers" will get you "better results." It does technically but the ultimate thing you need is the result, not just the AO pass. If in any case, I usually don't use AO, I fake it in the Filter Editor as much as I can using various techniques, sometimes results are better... sometimes it's nicer using AO. That's totally up to you to experiment.

You may also be rendering unreasonably huge resolutions. Alright this doesn't go for photo filters but it does go for textures. If you're rendering a seamlessly tillable texture, you probably don't need it at a 4k resolution unless there's a 100% need for that resolution. The purpose of seamlessness to begin with is to make it easier for yourself to handle surface wrapping and making smaller files. I think 2k resolution is quite good enough but if you need more, do go ahead and do it. I've seen interesting threads on this forum regarding people rendering 10k resolutions. I don't know why you'd need such a ridiculously huge resolution, but that's cool in its own way. As for the general user, think about what you're doing and what you really need. Don't waste time on huge resolutions "just because you can."

Think about what you need to render out. By default a surface filter will render out every pass, so the surface looks 3D in FilterForge. But for preview purposes, maybe you can just look at the diffuse layer and determine if you like the color or not. Look at the bump layer separately and normals separately. Then once you like what you have, see what it finally looks like. A lot of the time, I myself just render out the surface filter as is, but in some cases, I'll switch over to the diffuse layer for preview purposes if the filter is a little sluggish.

Get a little cleaver: design your work so it doesn't have to always rely on a filter. Do note you can generate nice textures using programs like Photoshop to a degree, and then use FilterForge and enhance what you got. Basically spare your time a bit by doing something manually and then make FilterForge do the rest.

If you're making a filter: read that awesome wiki on Dos and Don'ts of FilterForge. It saves you time and also your render time. Utilize techniques and avoid creating round-about ways of making effects, organize your filters and structure it in a manner that allows for optimal and at the same time expandable outputs.

Most people don't think about stuff unless they look it up. So as much as we can ask for a faster render engine, it also boils down to how we deal with it and how we use it. Be smart, think about what you're making, not about what the software is lim ited to.

Good lecture, Skybase
Now let's make it a debate

Quote
...the current progressive render in FF3 is quite fast. It gives you visual feedback relatively quickly...

Most of the time that is true. You can look at the first row being rendered and see if you like where things are going. It's easy to change the settings and let it start again. Even if sometimes what you really wanna see is closer to the middle part of the filter, you can just zoom in to the max on the place you want to see, and it will render first. Then you can zoom out to see it in real size and decide if that's what you want.
This is a really cool feature that FF has, that I haven't seen in a lot of other programs.
There is the occasional filter that takes a long while untill it starts rendering, and there's nothing you can do about that. I don't know the reason why some filters need a lot of time to "prepare themselves" before rendering, but as far as I know, there aren't many that do so.

Quote
Let me mention a couple things that slow your render down. First of all, depending on anti-alias settings, you can really get slow results.

It's true that AA can take a long time, sometimes longer than the render itself. The AA pass, however, starts only after the filter has done rendering, so you should already have a good idea if the result is to your liking or not. AA rarely changes the end result SO MUCH that you really have to wait for it to decide if you need to change some settings. Again, zooming to the max on a specific area will make Anti-aliasing start there first. So even if AA would make a dramatic impart on your final result, you can test it out on a few key areas instead of waiting for it to go over the whole image.

Quote
Don't assume "high numbers" will get you "better results."

That's true in many aspects, not just Ambient Occlusion. There's a limit to what the human eye can see, so you don't have to OVER DO something and make FF work for nothing. If you wanna kill a fish, it doesn't matter if you drop a 1 ton rock on it's head or a 500 ton rock. It won't get more dead. It would however take you much more effort to lift the 500 tons rock.

Quote
The purpose of seamlessness to begin with is to make it easier for yourself to handle surface wrapping...

True. The only reason to use a TILEABLE texture is so you don't have to cover the entire surface with exclusive data. Many times you can use a relative small texture for a large surface without it being too obvious. As long as there's not TOO MANY repetitions of the same feature on the surface, nobody's gonna notice.

Quote
As for the general user, think about what you're doing and what you really need. Don't waste time on huge resolutions "just because you can."

The fish rule applies here too.

Quote
for preview purposes, maybe you can just look at the diffuse layer and determine if you like the color or not. Look at the bump layer separately and normals separately.

Right. The more you work with bump and normal maps, it's easier for you to imagine how they all work togather as a 3D surface. Think of it like a blind man reading braille. The information is in the texture. You need to "feel" the bump and normal maps. It comes with practice.

Quote
design your work so it doesn't have to always rely on a filter...

While I agree it's never a good thing to rely exclusively on one tool, and that much of the work can be done manualy BEFORE using FF, I bought the tool for using it Think of it as a farmer buying an expensive harvester. He should use it as much as possible, but he must never forget how to use the sickle.

Quote
If you're making a filter: read that awesome wiki on Dos and Don'ts of FilterForge.

YES! Learn from other people's experience. The wheel only needs to be invented once. Don't try to rediscover fire - we all know about it. Use the knowledge of your FF ancestors

Quote
Be smart, think about what you're making, not about what the software is lim ited to.

Good advice for life, regardless of FF


Now I will add something of my own. If you're unsure about what you're looking for, use the reduced size preview rendering. That way you will be able to test more options in less time. However, If you know exactly what you want, disable the reduced size render, and render at full size. That way you will avoid having to Re-render when you save.

Quote
Morgantao wrote: The only other software that does what FF does costs around $1000. It's ment for professinal use at game\animation studios, and it's more robust and advanced than FF. Even the interface is more complex. A true PROFFESIONAL product.

I agree that the other software that I suposse (I relly do not know much) makes about the same it cost 1000$ AND the filters are NOT FREE as it happens in FF. But as you say very well, this software is specially done for PROFESSIONAL companies and gives them what they need, and would be impossible that it would be slow and have problems like FF have. As you say this is TRUE professional product.

I do not think that this is the ONLY software that can be compared to FF because Genetica also works in some way (although the uses are different) like FF so this could be also used, and Genetica IS USED ALSO BY PROFESSIONAL COMPANIES, Why because it does not have the speed problems that FF have and is more oriented to professionals, and it cost the same 400$ as FF.

Quote
FF however is more of a hobbyist oriented. It's for everyday ordinary people. Amatures even. That's why the price is much lower. That's why the render engine is 6 years old. That's why it's not 64bit. That's why it can't use more memory. That's why simple feature requests never get implemented.

YES I agree with you in the points above

FF is more of a hobbyist oriented BUT NOT because probably is the FF Inc. intention to be this way, is ONLY because is NOT up to the standards and requirements that the PROFESSIONAL companies and user need and ALREADY HAVE from other software companies.

Would you ask to a professional company to wait 23 minutes to render a 4000x4000 image texture ? SURELY NOT IN ANY WAY

Quote
Morgantao wrote: However, you can rebuild pretty much any FF filter you want in the other software, and see it in realtime.

You could do that ????

Does the Substance Designer have the same filter editor components as Filter Forge ???

I will download the trial and see it.

Quote

You may also be rendering unreasonably huge resolutions. Alright this doesn't go for photo filters but it does go for textures. If you're rendering a seamlessly tillable texture, you probably don't need it at a 4k resolution unless there's a 100% need for that resolution. The purpose of seamlessness to begin with is to make it easier for yourself to handle surface wrapping and making smaller files. I think 2k resolution is quite good enough but if you need more, do go ahead and do it. I've seen interesting threads on this forum regarding people rendering 10k resolutions. I don't know why you'd need such a ridiculously huge resolution, but that's cool in its own way. As for the general user, think about what you're doing and what you really need. Don't waste time on huge resolutions "just because you can."

Our customers use our 3,600 x 3,600 tiles (file sizes range fr om 5 MB to 15 MB) to fill areas ranging from a couple of feet to 40 feet or more. The products they are producing, such as banners and vehicle wraps, are big. Their main concern in using seamless tiles is not ending up with a noticeable pattern in the rendered graphic. A repetitive pattern is noticeable to the trained eye very quickly, even with just two tiles. It will become obvious to even the untrained eye for most tiles with four or more tiles in a row. It is, therefore, a huge need for us to supply tiles that fill a reasonable area in order for the client to find our images acceptable.

So taking that as a given, it becomes obvious that, for us at least, FF is far too slow. We would have no great complaint if the slowness was lim ited to the rendering of the output image. But that is not the case. The slowness is also a major issue with the preview image as well. This multiplies the 30 to 60 minute rendering time to several hours when you add in the waiting for variations to display when we tweak for the least noticeable repeat pattern.

One answer for FF would be to adopt a solution such as provided by Genetica with their smaller preview display. In their separate window, the user can set it for 256 x 256 through 1,024 x 1,024 and get an acceptable preview of both the image itself and how it will look when tiled any number of times. And the preview works in an instant instead of taking half a minute to several minutes to show every tweaked change.

Regretably after the answer of GMM, that confirms that FF will not have GPU acceleration, and have not confirmed or given any news about any update to the FF render engine to make it faster, the answer to the question on the title is:

NOTHING

I agree that SD is much more comples and much less user friendly, but trust me, once you feel comfortable making complex filters in FF, the transition to SD will look less and less daunting.
You and I, right now, don't need to use SD, as it's mostly dedicated to profesionals and big companies. But if FF won't do something drastic about the render speed in the near future, great filter makers will start making the tranitions to SD, which will affect the quality of new filters for FF, which will turn away more users...

My Machines that i Build, are water Cooled, And OverClocked, 5.2Ghz. Intel I7 980 Extreme..

Ok Next machine will be a Xeon Based.. Dual Socket..

There fun to build.. Computers on Steroids.. yep yep

Just When I render out 32000x32000 res it does take a little while..
Hour and a half..
So I will to the big stuff on my other work station..

Quote
3DCGMODELER My Machines that i Build, are water Cooled, And OverClocked, 5.2Ghz. Intel I7 980 Extreme.. Ok Next machine will be a Xeon Based.. Dual Socket.. There fun to build.. Computers on Steroids.. yep yep

Do your work selling this computers, or you do you build them for yourself or someone else?

As Skybase said, the only reason I see for a 32000 x 32000 resolution is very high quality large scale print

BY the way the XEON is expensive so Intel have released the new Sandy Bridge E instead that is cheaper and with more perfomance than normal desktop chips and more similar to the Xeon one but with 6 cores instead of 8

My feeling is that the member created filter library is both a blessing and a curse. A blessing because it gives all licensees a great starting point and for FF a great selling point. It is a curse because there are so many filters that are not optimized for speed better than they are. One is left to wonder what the library would be like if FF took the time to inspect each filter and to apply some quality control and modification to those that need it.

Why not render at say half or a third the resolution you need and then use a program like Alien Skin Blowup or BenVista PhotoZoom?

Quote
Well... I guess we need a premium library then... ohhh waiiit. smile:p

That would be fine by me but since there isn't one the question remains: Why doesn't FF perform QC on what is one of their biggest assets?

Quote
SpaceRay wrote: ..it cost 1000$ AND the filters are NOT FREE as it happens in FF.

Well, now since 1 June 2012, costs 590$

Quote
Morgantao wrote: Correct, the software is $1000 and the filters cost extra. However, you can rebuild pretty much any FF filter you want in the other software, and see it in realtime. Also, you can have as many custom folders as you want, and arrange the filters however you want

HOW COULD YOU DO THAT?

Quote
SpaceRay, while not trying to do so myself, there's another thread on the forums that talkes about substance designer, where some person said he uses SD at work because of the FF library clutter, and that he just recreates the filters he needs in SD. SD does have a node based filter editor, it's just more intimidating to use.

Please can you tell me WHO is the one that is able to recreate FF filters in SD?

Is there ANOTHER thread where is this information different than this thread Has anyone here used Allegorithmic's Substance Designer?

Quote
Please can you tell me WHO is the one that is able to recreate FF filters in SD

I can't seem to find the thread where this was discused, but someone talked about being in the process of migrating all his filters from FF to SD, because his work place won't get FF for memory and speed issues.

Quote
if it could be possible to transfer in some way the filters from FF to SD

Yes it is possible, just like writing the whole bible in your own hand writing.

Quote
even I f would need to rebuild them in SD

There's no other way.

Morgantao, of course that there is no other way than rebuilding the filters, because of course that there is NO possible way to import or any software to convert the filters other than by hand, AND also very important is that they DO NOT have the same components.

BUT the company of Substance Designer (SD) have CARED to make this software faster and optimize it as much as they can so they can offer it and be attractive to professional companies that value much the time spent to get the results and things done.

SO I would prefer to spend the time once to convert the most used and best filters to SD and be able to use them in a much better and faster environment and not keep waiting more than 15 or 20 minutes ( or more) for a 5000 x 5000 result fr om a complex filter.

Quote
Skybase wrote: I highly doubt it SpaceRay. I mean... that's just based on almost nothing.There's barely a remark regarding the status of the new version. No press release, barely any commentary.

I do not want to go against you in anyway, but this has been requested since 5 years ago and still there is no news, comment, remark or anything that will tell you that this will happen, the only thing is that they are blocked by the way FF was built and programmed in the beginning with it´s own custom memory manager and the way FF processed things.

Sorry but my commentary is based on some other commentaries from GMM that was saying that they were not going to optimized the code engine or the render engine because they needed to rebuild the whole thing and that was TOO MUCH WORK and they will not do it

Also after I asked, GMM also confirmed that FF will not have any kind of GPU acceleration, so if you join both you get that FF will not get any kind of speed increase and I suposse that FF 4.0 will have the same speed of FF 3.0 or slightly higher.

Perhaps you have missed others comment on others threads (that I do not remember wh ere now) that has the reason and what is based MY comments and WHY I know that FF was NOT going to make a faster FF.

Quote
Skybase wrote: Though over the past 3 versions the speed's increased for every upgrade. It's kinda expected of anyway.

As far as I know from the forum AND testing myself the speed, there has been only 1 speed increase from FF 1.0 to FF 2.0. Then when FF 3.0 was in beta there was much requests to make it faster than FF 2.0, but when the final version was released it had the same speed than FF 2.0, I have already posted about this and made some real world tests using the same filter in both version and got around the same render time.

Quote
Skybase wrote: It's kinda expected of anyway.

OH YES! It has already been VERY expected since 5 years ago, and since then the only speed increase (that I can´t confirm how much it was, and think that not much) was form FF 1.0 to FF 2.0

And as FF wants to remain silence and not make any news about if they are working on any kind of speed optimization I have decided NOT to keep asking for it because is tiresome to keep 4 years requesting something and FF has done nothing for it yet, and have given reasons WHY this has not happened, AND WHY will not be possible to happen in the near future because it would require a whole rebuild of the core engine or the software and they have said that they will not do it.

And also I will not keep requesting this and will not depend on what FF Team does, and will look for alternative POSSIBLE ways to make FF faster like buying a new powerful and faster DUAL HEXACORE CPU computer with will surley make FF faster or use an alternative software.

Quote
Skybase wrote: basically means that's just a faulty assumption based on previous assumptions. Saying things too early in the game.

Is not previous assumptions, as I have said there are already real facts that FF team and developers have tried to make a speed increase and they failed to do it because it need much work to do it and prefer to concentrate on things that are more easier and possible to do.

Look everybody... ART!

 

More of the art I have been making (note this are 4 ones only joined together here for the forum, they are separate ones)

Quote

Sign Guy wrote: My feeling is that the member created filter library is both a blessing and a curse. A blessing because it gives all licensees a great starting point and for FF a great selling point. It is a curse because there are so many filters that are not optimized for speed better than they are. One is left to wonder what the library would be like if FF took the time to inspect each filter and to apply some quality control and modification to those that need it.

This is so very true, and sadly common with any product that leverages UGC (user generated content). About 99% of the filters in the library are completely useless, many are pure nonsense. For example "Electrical Outlet". Has no useful value in this context. Also many are just repetitious. I was looking for a quick brick pattern the other day and thought I would use FF. I search for bricks and get 276 results. Many of which have nothing to do with bricks.

What compounds this issue is that the sorting/filtering capability of the filter library is practically non-existant. Ranking/rating would be helpful. Some moderation of keywords would be useful. Anything to help wade through the sea of crap to find something useful. Maybe a "featured" section where FF could promote quality filters or something like that.

While being faster would be certainly be nice, ultimately, FF is pretty much a toy that is useful on occasion. Making filters easier to find, better interface, and expanded features are much more important than speed.

ZG

There is the problem biggest problem with FF. 99% are useless for anyone using it in any professional capacity. Like the filter I mentioned (there are hundred more), any professional doesn't need a 3d party photoshop filter for an electrical plug. Additionally, they probably wouldn't even considering looking there. So that leaves hobbyists. Sure it may be useful for someone in that category, but then speed isn't a concern.

And there is the heart of it. FF isn't a serious tool professionals rely on, not because of speed, but usefulness. Pros are not their market. So I can't imagine that speed is a serious concern to the developers of FF. At least not one to spend resources on. Boosting the speed won't make it a pro tool. And the hobbyists who play with it aren't going to be turned away because it is slow. Investing in more fun features is where FF is going draw more of its base of users.

--

From the description, it appears FFCat is a windows app. Interesting, but I don't do windows. And while useful, being an unofficial solution it can't be relied on. Should FF block it, or change it's structure or you are unable to update, it could easily no longer be available. Nice work though. But FF should have something like this as part of the package. Maybe they could license FFCat from you.

--

Featured section isn't sortable and still contains a lot of silliness. Too broad. The people looking for 3d textures, are not the same as the people would would use it for photo manip, and not the same who would use the cute little whale "filter". It needs featured per category and an advanced search.

Of course everybody wants more speed, I would never suggest they don't.

I was just saying that for their target audience, making it faster won't impact sales very much if at all. At least probably not enough to justify the development cost. For the hobbyist, more slick, shiny, fun features will increase sales.

And for pros, speed really wouldn't have an impact on sales either, because speed isn't the biggest problem.

I am a pro (games), and got it when it first came out. I upgraded once, but have no interest in upgrading to 3. It has come in handy a few times, mostly for cleaning up pencil drawings. But it really doesn't have anything that I can't do by hand. And it is becoming more difficult to find anything of use.

It would be great if it were faster, but I just can't see the developers putting resources to performance over features.

My assumption that its primary market (currently) is hobbyists is a combination of the content of the forums and a bulk of the filters themselves. I know of virtually no other pros that use it, and most have never heard of it.

Actually, FF is not a really a plug-in for PS. PS is not required to use FF at all. It is a stand alone app. In fact if you do use it with PS, all is does is launch FF and return the results to PS. It can be used just as effectively fr om Gimp/PSP/nothing... with just a couple of extra steps. In fact, because of the speed, on the rare occasions that I do use FF, I don't use it from PS. If I launch it separately, I can render and still use PS for other things. Running it as a filter holds PS hostage while it renders.

And again, to be clear, I am not saying in anyway that speed isn't important. I am saying that I can understand why the creators of FF may not find performance optimization a valuable investment of resources. Speed optimization would most like involve a ground up rewrite of the core engine/tech. It isn't simply a fix.

Everything else you wrote is on the money. I purchased because it was a collection of quick filters that could save time painting. Honestly, since the first version, FF has offered nothing new that would entice me to continue to upgrade.

I have played with the filter editor, but really don't have a need to actually use it. If I need a texture or effect, I generally just need that instance, not the ability to create it in the future with variation. It is often faster (if a filter doesn't exist) to just create that texture in PS/AI.

I think "missing the boat" sums up FF's strategy very well. I get the feeling they had a cool idea for a way to create filters, but not really sure how to market position themselves. And by not being clear about who they are marketing to, they end up not being good enough for pros, and maybe a bit tech heavy for casual users.

When it first came out, I played with the demo for under 10 minutes before I bought it. It seemed like it had great potential and was excited about wh ere it would go and what filters would come. Today, it is still pretty much that same app I bought several years back. A quick way to to whip common textures quickly. In that respect, it is still worth every penny I spent on it, and glad I have it. But it's lack of performance, lack of direction and useful new features, leads to two things for me:

1) no reason to upgrade, regardless of sale prices, and
2) it is getting replaced in my tool kit by competitors (and Adobe itself), who have stepped up to fill in the spaces it use to be unique to FF.

Heck they are missing boat in huge other ways as well. Someone mentioned Substance Designer. SD is a simply amazing tool. Virtually nothing in its class. (heh... my company is listed on the SD product page) However, most of the time you have to start with some sort of bitmap texture, to build your substance. I have used FF for this. FF would be an awesome complement to SD as either a plug in of sorts. or simply marketing their tool as a great way to facilitate substances.

Hopefully they will figure it out, boost performance, and find their place. I still feel it has potential, but is slowly becoming a "could have been".

ZG