A trick for saving render time

[Trash_Panda]

Rendering models only takes a lot lesser time than rendering a whole scene with models in it, i think we all are at the same page in here when it comes to that.

so...

...with that in mind, i thought of something like this:

1. We prepare the scene with models in it, we hide the models and render the image.
2. We hide the scene and render the models.
3. Copy&Paste
4. Profit??

At this point ya'll are saying what is the point, maybe some dont who understood where im getting at. The point of doing this is giving different expressions and poses to characters without rendering the whole scene every damn time. By doing this you will still render the models every time of course but there is still the time save by excluding the scene from renders, the background will look like a still image.

I know this is already a widely used thing especially in RPGM games, nobody is reinventing the wheel. Here's my problem, you cant preserve lightning with this method. So if you're a user of this method, could you guide us of how you do the lightnings correctly?

Thanks in advance for any help.

 

[rayminator]

I do this each time I have more then two character in the scene but I do it differently

I apply the scene and then merge the models and then save then remove the models then render then close daz to free vram from the video card and then reload the save and delete the scene and then render again

 

[raventai]

Hi, this is what I do in Studio for complex scenes :
prepare the scenery (extreme cautious with sky, shadow, lights, obstructions, reflective materials...).
Render a scene with the final lighting (my background) and cam (the same for all the process). Save it !
Load and pose characters (minimum one maximum 5, you can try up to 7 depending on your rig)
NOW, save each character as a scene subset with the lighting (capital) and cam, you lose some reflective material but not seriously
then you can empty your scene and reload a new empty one, then you load each subset and render it (can be very fast).
After that your work in photoshop layers of course.... The difficult part is posing in the right place guessing the shadows and to have a very efficient and mastered set of lights for the characters. Extremely heavy scenes can be done that way. Now it is my way and others have probably more efficient processes.... a last word : you cannot remove parts of the scenery or characters before rendering, the memory is lost due to DAZ poor progamming....

 

[Trash_Panda]

Thanks but these still cant solve my problem
Ok
Lets say you're doing our famous park scene, its easy to get the lightning and shadows done since you're open in the field anyway, you can just create an empty scene with your character and add lightning to sky, done.
However in closed places like train, bus or office you cant exactly use the same method and capture the perfect lightning and shadows. Maybe in office you can just try to simulate lamp lightning and get some nearly perfect output but when you're in a bus in daylight, you gotta at least include the bus itself to simulate similar lightning which increase the render time anyway and we want to avoid that.

My question more or less is if there is an option in daz that you can somehow preserve the lightning and shadows even when you hide the scene, or you keep the scene but render without it... something like dat.

 

[Deleted member 1121028]

Rendering models only takes a lot lesser time than rendering a whole scene with models in it, i think we all are at the same page in here when it comes to that.

so...

...with that in mind, i thought of something like this:

1. We prepare the scene with models in it, we hide the models and render the image.
2. We hide the scene and render the models.
3. Copy&Paste
4. Profit??

At this point ya'll are saying what is the point, maybe some dont who understood where im getting at. The point of doing this is giving different expressions and poses to characters without rendering the whole scene every damn time. By doing this you will still render the models every time of course but there is still the time save by excluding the scene from renders, the background will look like a still image.

I know this is already a widely used thing especially in RPGM games, nobody is reinventing the wheel. Here's my problem, you cant preserve lightning with this method. So if you're a user of this method, could you guide us of how you do the lightnings correctly?

Thanks in advance for any help.

What you're gonna want to do is possible but ain't gonna reduce rendering time. Some part of your render will be just rendered as alpha, but the whole scene will still be rendered for accurate shadows.

Select your node in daz(your figure for ex)>Click on it, select/select Children
Render setting>Canvas>Tick "canvases">Click "+">Tick "alpha"
Click nodes>Create from selection> "Yes"> "accept"
Hit render.

edit:

You must be registered to see the links

your real rendering time crusher. Master it and it's ezpz.

 

[MissFortune]

At this point ya'll are saying what is the point, maybe some dont who understood where im getting at. The point of doing this is giving different expressions and poses to characters without rendering the whole scene every damn time.

Couldn't you just spot-render the character with each new pose/expression? You wouldn't have to adjust lighting or much else. Just have to make sure they're in (relatively) the same position.

 

[Rich]

However in closed places like train, bus or office you cant exactly use the same method and capture the perfect lightning and shadows. Maybe in office you can just try to simulate lamp lightning and get some nearly perfect output but when you're in a bus in daylight, you gotta at least include the bus itself to simulate similar lightning which increase the render time anyway and we want to avoid that.

So, one way you can tackle this is by rendering your train, bus, whatever into an HDRI, and then render your characters using just the HDRI. This gets the lighting on the characters to be "correct" based on the surroundings without requiring you to re-render the surroundings each time. The idea is that you create a "dome render" centered at the character which captures a 360-degree of the world as seen from the character's POV, then turn it into an HDRI.

Of course, if you do this, your characters won't cast shadows onto the background unless you add a "shadow catcher" to simulate where the surface they would be shadowing onto would be. But, depending on the application, this may not be a big deal.

The "render your background as an HDRI" approach works well if your camera doesn't move from character render to character render. If you start moving the camera around, you'll see some distortions in the background.

Here are some links about creating your own HDRI's:

You must be registered to see the links

You must be registered to see the links

You must be registered to see the links

And here's a video on iRay shadow catchers

You must be registered to see the links

 

[Mikethe3DGuy]

FWIW, this used to be a major concern of mine when I was running my previous GPU. Since upgrading to a 3090 the amount of time I'd save by setting these kinds of workarounds up (and then the required postwork after) is more than negated by faster render times with the new card. And in fact, usually when I had to go through steps like these, it wasn't just to reduce render time but because I didn't have enough VRAM to even render the whole scene in one pass. I know not everyone has the freedom to upgrade like that but I found a way because I was simply exhausted wasting time and brain cells trying to solve problems like these.

 

[Trash_Panda]

FWIW, this used to be a major concern of mine when I was running my previous GPU. Since upgrading to a 3090 the amount of time I'd save by setting these kinds of workarounds up (and then the required postwork after) is more than negated by faster render times with the new card. And in fact, usually when I had to go through steps like these, it wasn't just to reduce render time but because I didn't have enough VRAM to even render the whole scene in one pass. I know not everyone has the freedom to upgrade like that but I found a way because I was simply exhausted wasting time and brain cells trying to solve problems like these.

Three things:
1-) Using a camera mod which removed all invisible background.
^ Basically costs 0 time.
2-) Using scene optimizer.
^ You only do it once for a scene, even then it takes 5 minutes at worst, at least from my experience.
3-) Remove anything thats outside of camera
^ Costs like uhhhh... 5 to 10 seconds?

So yeah. With DAZ there's no perfect scene as convergence ratio can never be %100 iirc. My aim is getting around 2000 iterations for every scene with 1920x1080 resolution, result is usually satistifying enough. As long as small details are detailed enough, its sufficient for me. My limiting factor is time for each render. 3 hours always give good enough image quality.

Btw i just noticed how i came a long way with daz since i started this thread lul.

 

[Mikethe3DGuy]

Three things:
1-) Using a camera mod which removed all invisible background.
^ Basically costs 0 time.
2-) Using scene optimizer.
^ You only do it once for a scene, even then it takes 5 minutes at worst, at least from my experience.
3-) Remove anything thats outside of camera
^ Costs like uhhhh... 5 to 10 seconds?

I use Scene Optimizer, but honestly I've found it takes much longer than 5 minutes, especially since after optimizing you have to close Daz as per SO instructions, wait the 10-20min (in my case) for Daz to deal with its well-known memory issues before it can be re-opened, then re-open Daz, and re-open the scene.

But I'm referring mostly to the whole concept of separately rendering figures and background then figuring out how to get figure shadows to fall on the scene, especially when doing indoor scenes with photometrics and emissives (almost all my scenes are indoor)... or alternatively turning your BG into an HDRI.

 

[Droid Productions]

Here's my problem, you cant preserve lightning with this method. So if you're a user of this method, could you guide us of how you do the lightnings correctly?

Set up a camera where you want to sample the light from.
Enable fisheye lens mode
Render out a single image (it doesn't need to be ultra-high resolution, we'll just be using it for lighting)
Take the image back in, set it as the HDRI, make sure rotation's right.
Disable rendering background (we'll post-composite the background you've already rendered; HDRI backgrounds for anything close never looks very good, even at ridiculous resolutions).

Some discussions here:

You must be registered to see the links

 

[Trash_Panda]

Set up a camera where you want to sample the light from.
Enable fisheye lens mode
Render out a single image (it doesn't need to be ultra-high resolution, we'll just be using it for lighting)
Take the image back in, set it as the HDRI, make sure rotation's right.
Disable rendering background (we'll post-composite the background you've already rendered; HDRI backgrounds for anything close never looks very good, even at ridiculous resolutions).

Some discussions here:

You must be registered to see the links

Rich up here basically shared the youtube link of what you just said.

I use Scene Optimizer, but honestly I've found it takes much longer than 5 minutes, especially since after optimizing you have to close Daz as per SO instructions, wait the 10-20min (in my case) for Daz to deal with its well-known memory issues before it can be re-opened, then re-open Daz, and re-open the scene.

But I'm referring mostly to the whole concept of separately rendering figures and background then figuring out how to get figure shadows to fall on the scene, especially when doing indoor scenes with photometrics and emissives (almost all my scenes are indoor)... or alternatively turning your BG into an HDRI.

Idk man, if im going to use the scene at least 4 or 5 times (which is always in my case) then that 20 minutes can very easily be overlooked. Ofc you might not want to do it now with your 3090 gpu.

I've given up the idea of rendering seperately using HDRI backgrounds, i've already wasted two days trying to work that out and its just not worth it.

 

[Mikethe3DGuy]

Idk man, if im going to use the scene at least 4 or 5 times (which is always in my case) then that 20 minutes can very easily be overlooked. Ofc you might not want to do it now with your 3090 gpu.

I've given up the idea of rendering seperately using HDRI backgrounds, i've already wasted two days trying to work that out and its just not worth it.

That's when I tend to use it too, but yeah not as often with the 3090. Can be difficult to keep all the texture names straight though. Sometimes I'll accidentally allow a texture reduction that turns out to be a foreground item and have to go back in and manually revert it.

I might start experimenting with custom HDRIs for lighting only.