Messages

31

[Max] I need help! / Re: Perforated Brick

« on: 2018-01-18, 16:44:56 »

Thank you all for your input! And thanks for the link to the model. However I will need to add the mortar nbetween the bricks, which I can't find a easily modifiable way of modelling this so i can control and edit the brick and mortar elements. I am also looking for a white brick texture, do people use the corona materials website for there textures? I have Siger Shader installed but no such high quality brick materials.

THanks again for all the help. This forum is pretty awesome and have only just started using it.

You could model the mortar around a single brick and then scatter/array them together as you need :) Shouldn't be that much of a problem to be honest.

EDIT: Added a simple example - If you need more detail or combine it with non-perforated brick wall, you can add mortar all the way around and scatter away :)

One quick improvement would be to attach the mortar mesh to the BOTTOM of the brick so that you don't see it at the top xD

32

[Max] I need help! / Re: Sampling issues

« on: 2018-01-18, 14:05:15 »

I think this issue is similar to LightMix producing noise if specific light is completely turned off or increased a lot. So the beauty pass is mainly considered for sampling, and it does not really matter what is "underneath" some clean area (like that white spot in your examples). I would say this is expected as a result of using adaptivity.

Just a note - there are 3 things to distinguish:
1) The new light solver - exposed to the user at the top of performance tab, off by default, affects scenes when rendering many lights/complex lighting scenarios, our inhouse tech
2) The new random sampler - hidden in the devel/debug rollout, on by default, generally affects speed of noise refinement in the rendering, Chaos Group tech (the only in Corona right now)
3) The new enviro sampler - hidden in the devel/debug rollout, on by default,  affects noise mainly in scenes lit by HDRI, our inhouse tech

In my case, as already mentioned, it was the adaptivity. I turned that off and the AO pass rendered as normal and it's the first thing I tried - it's pretty obvious what the problem is.

My question is whether it's possible to have another way of sampling render passes later down the line while keeping the adaptivity for the beauty pass?

33

[Max] I need help! / Re: Sampling issues

« on: 2018-01-18, 11:07:25 »

I turned off adaptivity in the debug settings and that sorted the AO sampling problem - it's sampled as I would expect. Though, of course, the beauty pass is taking a bit longer to clear as a result.
Is this something to consider a bug or is it a "feature" we'll have to live with every time render passes are needed?

34

[Max] I need help! / Re: Sampling issues

« on: 2018-01-18, 10:09:30 »

I'll have to try that and see what happens. I left one image to render overnight on the new sampler and it did clear up after 16 hours xD

35

[Max] I need help! / Sampling issues

« on: 2018-01-17, 18:09:07 »

Hi, guys!

I've got a question about sampling. I think it's since the last update here VRay's adaptive sampling was introduced I started getting really crappy sampling in AO and certain other passes, which annoyed me on this latest job especially.

If you look at the images below - I think it's because the sampler just sees the white reflection as finished and doesn't give it any more samples (look at where the white reflection ends on the right side of both passes!) - which is fair for the beauty pass, but the same areas are also neglected in other passes, such as AO, where it makes no sense and makes for a very blotchy render with noise, which is quite annoying... Happens all the time now and I hate it because resolving those areas either requires manual intervention or leaving the image to render forever - neither of which is viable when rendering a job in bulk, say, via Backburner - and I have one job coming up where I'll have 300+ renders to make of various parts for a company.

Is there anything I can do about it beside having to focus rendering power on those areas every time I need an AO pass (area masks)???

Thanks!

36

Hardware / Re: xeon v4 cpus

« on: 2018-01-16, 10:14:48 »

I know this is a wee bit off-topic, but I'd much rather prefer to work with top-end i7 chips than Xeons... Had a 20-core Xeon machine - it was slow in single-threaded tasks (say....MAX viewport!) at 2.6GHz and got superseded by a heavily overclocked 8-core i7-5960X (from 3.2GHz all-core to 4.5GHz all-core, on water). cheaper, much more responsive and a pleasure to extract more power from by overclocking. You do lose out on being able to have hundreds of gigs of RAM and having more than one socket, but for a workstation I wouldn't ever buy a Xeon anymore...at least not with those pathetic clocks...

Latest gens of Xeons can overclock single/two/four cores up to 3.6+ Ghz, they don't suffer from slow single-thread tasks at all.

And previously, i7s were much slower in multithreaded, making them poor choice for people who do a lot of quick test renders and interactive. Having monster machine can really speed up look development workflow in cpu renderers.

But right now I would agree. The best workstation right now is overclocked i9 7980XE. Basically Xeon machine.

3.6 sounds slow to me, considering my grandpa Haswell-E chip can handle 4.6-4.7 at the same core amount, albeit the IPC is slower than the new chips. I think I saw that i9 handle 4.8GHz on all cores somewhere, which is huge IMO.

37

[Max] I need help! / Re: Clay texture with finger prints

« on: 2018-01-15, 17:44:09 »

I'll be cheeky and plug the Corona tri-planar map. Been making all sorts of great "no UVs required" materials for recent jobs with it. Though I reckon won't be good for animation.

38

Hardware / Re: Motherboard recommendations for Threadripper?

« on: 2018-01-15, 17:40:26 »

I have 4 GPUs in my rig, 2 1080Ti FE and 2 1080 FE. I use them for mining nearly 24/7 (while I'm at work, at home relaxing and when sleeping) and I use them for FStorm GPU rendering - blazing fast stuff to have alongside Corona. Though I will say I prefer FE cards - if anyone knows how to make a decent board - it's Nvidia. Far too many failed partner boards out there for my liking, so I'll sacrifice a bit of cooling for reliability. Had all 4 cards since launch and they're kicking like new. Getting to 2100MHz core on all 4 cards isn't an issue under full rendering load either, which I can't say for many partner cards :)

39

Hardware / Re: xeon v4 cpus

« on: 2018-01-15, 17:36:37 »

I know this is a wee bit off-topic, but I'd much rather prefer to work with top-end i7 chips than Xeons... Had a 20-core Xeon machine - it was slow in single-threaded tasks (say....MAX viewport!) at 2.6GHz and got superseded by a heavily overclocked 8-core i7-5960X (from 3.2GHz all-core to 4.5GHz all-core, on water). cheaper, much more responsive and a pleasure to extract more power from by overclocking. You do lose out on being able to have hundreds of gigs of RAM and having more than one socket, but for a workstation I wouldn't ever buy a Xeon anymore...at least not with those pathetic clocks...

40

Hardware / Re: Monitor Calibration

« on: 2018-01-15, 17:27:12 »

Agreed with Juraj on the above.

I bought an X-Rite i-1 calibrator and the Color Passport for my photography needs and I have to say - for me it's worth having a calibrated screen just for the reasons of my colors being correct (my screens were either too green or too blue before calibration). It doesn't solve ALL the problems, though... Different screens still look different, even if calibrated to the same color space (though I guess in my case it's down to one being IPS and the other TN with different brightnesses). I've also had clients come back asking to adjust colors of images where, in face, we later found out that their cheap laptop displays are horribly blue and dark in comparison to a calibrated display.
Phones will also display utter over-saturated garbage, unless you have the option to use the sRGB color space on it (which most people won't be doing anyway, since it requires digging into the settings).

I've come to realize that unless you're working with high-end agencies/clients (we're talking Apple or Samsung or Canon) - your beautiful calibrated colors are for you and only YOU to see... I was surprised that here in the UK most printing companies have no clue what color standards they use (should be ISO FOGRA39), if any.

It's a can of worms...

41

[Max] I need help! / Re: Vector masks - is it possible?

« on: 2018-01-15, 17:18:35 »

I don't know if my message was invisible or if people just don't read, but I've already established this exact workflow.

Hahaha, Wow you are exactly right, I'm a tool. I somehow totally missed that message you posted. My bad, hahah.

Haha, happened twice already with different people, so I figured maybe there was a posting error xD

42

[Max] I need help! / Re: Vector masks - is it possible?

« on: 2018-01-15, 09:19:23 »

Since we've already established that we can't use corona to make your vector masks....

I don't know about you, but I personally have found photoshop is better at making vector masks compared to using illustrator's live trace.

1. Select > color range > select your black or white in your mask> change fuzziness to 200

2. go to paths panel > Alt left click the "make work path from selection" button > set it 0.9 pixels

3. File > export paths to illustator

I'm curious to know if the workflow I've used before gives you a better result than you are currently getting! If it works out,  then you can do all the steps in a recorded action and automate batch all your masks.

I don't know if my message was invisible or if people just don't read, but I've already established this exact workflow. It's all there in post #7 xDD
Just that I need to do extra upscaling/downscaling steps and set the path conversion tolerance to 2px in Photoshop to prevent waviness :)

Usually illustrator does the job very well with smaller files, but it seems to add too much pre-processing/downscaling where totally unnecessary at higher resolutions, which makes it totally useless at that point.

43

[Max] I need help! / Re: Vector masks - is it possible?

« on: 2018-01-12, 13:29:55 »

You seem to have ignored what I wrote two messages up, mate xD

I've already got an automatic workflow. Will only take me a few minutes to vectorize a hundred of masks at good accuracy :)

Plus I'm in no place to tell client to change their infrastructure because someone somewhere thinks rasters are better, that would be daft xD

44

[Max] I need help! / Re: Vector masks - is it possible?

« on: 2018-01-12, 10:59:27 »

As I originally mentioned in the 1st post, the clients use a web system where they can pick SVG elements for various parts and recolor them on-the-fly in the browser. Useful for product selections where multiple colors have to be made. That way I only have to render and produce overlay files for details and shadows and they can do color changes for various elements themselves as and when new options become available and old ones are phased out. Saves me rendering hundreds of variations too. You can't exactly do that with raster files as quickly. Plus it saves on memory and download bandwidth.

45

[Max] I need help! / Re: Vector masks - is it possible?

« on: 2018-01-11, 11:58:10 »

Thanks for the ideas, guys, but these methods won't be accurate enough for my needs, I'm afraid. I literally need a mask outlined and filled, so random toon lines, even if vector, will require too much manual work to be a viable way to do it on a job like this (it's a volume job - will need to process hundreds of these masks!).
Also, MR has been discontinued and I use MAX 2018 - sure as hell am I not downgrading for one job :D

Apparently Illustrator uses a low-res proxy internally past a certain resolution to trace, which explains why increasing the image res gave me same/worse tracing results. Cropping in on detail gave me a perfect trace, but that means I'd need to split each image into 6-8 tiles just to trace, which means merging the vector mask tiles back into one piece somehow - again, not viable as a volume job.

The workflow I ended up with (and I hope it helps someone later on here) is as follows:

1) Render all the files in bulk using backburner at a fairly large res
2) Separate the masks into a folder and run that trough a neural network upscaler at 4x res. This one runs on my GPUs on CUDA, so takes very little time to upscale the images. But it's important to use this instead of Photoshop to keep sharp lines!
3) Batch action: Open the mask images in Photoshop, select white and make a mask. Convert the pixel mask into a path using a 2px tolerance (this is where the upscale was important - without a fairly large tolerance the path ends up wavy, but at a larger tolerance I'm losing detail. At 4x the res I can afford to round off to 2px and keep the detail!). Downscale the file back to 25%. Save the path as an .ai file. This is executed automatically in bulk once I've set up an action.
3) Action, but executed manually (Illustrator is retarded - no batch processing!): Open the .ai files in bulk in Illustrator, making sure to keep the original pixel proportions of the mask file (there's a tickbox - important so that the vector mask is in the same place as it should be when overlaying the raster renders!). An action is set up to select all lines, release compound path, then select all again and make the compound path (this is important because the automatic compound path is wrong at first), then select the resulting path and fill with black (as required by client), then save file. So for this I need to essentially open any of the bunch of tabs and go "Play action > CTRL+F4 (close tab) > repeat" - sounds bad, but would only take me a minute or two for 100 files!
4) Open all the files in Illustrator again and go scripts>SaveDocsAsSVG - all files get saved as SVG files in a given location.

Sounds like a bit of work, but it really isn't, surprisingly, just need to set up all the actions and run them. Sure beats processing hundreds of masks manually. Though it would be even faster if Corona could save a vector mask *nudge nudge, wink wink* :) (or if Illustrator wasn't retarded and using a low-res proxy to trace....)