best way to do animation (daz3d/renpny)

[Cohibozz]

Hi all! i've a question for my renpy VN

i've to do some scene in animate (daz3d) to use in my renpy game.
to do it fast i'm thinking to animate only the character and put AFTER in static scene.

I can do a MOVIE with only character and then put as layer in a bg scene? daz3d animate do a movie without background? it is possible?

or i've to do a animate scequence in stop frame image and i've to put all the image in photoshop to add bg image first and then make the video with virtualdub?

in any case i can't move camera in animation for this video if i've to put in a static bg right?

 

[anne O'nymous]

I can do a MOVIE with only character and then put as layer in a bg scene?

Yes but...

daz3d animate do a movie without background? it is possible?

I don't know, and that's the most important part.

 

[Rich]

I can do a MOVIE with only character and then put as layer in a bg scene? daz3d animate do a movie without background? it is possible?

Yes, this is possible, but involves a bit of work.

You must be registered to see the links

"The Movie displayable can also be used to define a movie sprite, which is a sprite that is backed by two movies. The primary movie provides the color of the sprite. A second movie, the mask movie, provides the alpha channel, with white being full opacity and black being full transparency. "

The release notes for 7.1.2 also say: "The

You must be registered to see the links

displayable now supports a new mode the color data and alpha mask data are placed side-by-side in the same file. This prevents issues where a main and mask movie could go out of sync. "

So, to do this, you'd have to render your image sequence as a series of PNG's, so that you have images with transparent backgrounds. Then, for each frame, you'd have to create an second frame that acts as the alpha channel. (White where the video is to be opaque, black where transparent.) Then you'd have to either join those side-by-side with the originals, or else use them to build a parallel movie.

There's probably a way to automate creating the alpha frames using some kind of Photoshop scripting, since what you're doing is just extracting the alpha channel from the PNG, but I'm not a good enough Photoshop guru to tell you how. (I'm thinking that this would actually be a cool tool to code - I know how I could do it in Java, for example.)

or i've to do a animate scequence in stop frame image and i've to put all the image in photoshop to add bg image first and then make the video with virtualdub?

There are ways this could be scripted using command line tools. For example, ImageMagick is capable of overlaying images on one another, so you could, in theory, build a .bat or .cmd file (if you're on Windows) that would work through all your images, overlaying them on the background and then spitting out a new set of images that could then be combined to produce the video. Similarly, I bet ffmpeg could do this kind of thing, but I haven't researched it. (I use ffmpeg to turn image sequences into movies, but I don't know every bell and whistle it supports.)

in any case i can't move camera in animation for this video if i've to put in a static bg right?

Well, you COULD move the camera, but this would probably create the effect of the character moving against the static background. In theory, that could be useful in some cases, but might be hard to get right.

 

[Cohibozz]

Tnks a lot for info

 

[khumak]

If you mean sex animations, IMO the biggest hurdle to making good ones is the fact that the human eye just seems hypersensitive to movement. So if you have an animation where parts of the figures are moving while other parts are completely motionless with respect to the rest of the figure, it just looks wierd. Unfortunately that means you either end up having to do close ups to reduce the amount of posing you have to do for each frame, or you need to manually change joint positions for just about every major joint in the figure in such a way that it still looks smooth.

The best solution I've found is to use something like one of the animation posing series done by Stimuli (

You must be registered to see the links

) and then you just need to manually tweak minor aspects of each frame like hand/finger position and genitals depending on what angle you're using.

 

[Rich]

If you mean sex animations, IMO the biggest hurdle to making good ones is the fact that the human eye just seems hypersensitive to movement.

This is basically true regardless of the animation - the eye is very sensitive to how things move in the real world, and it's difficult to create good animations that look realistic. It's not helped by the fact that DS is not well tuned to doing animation - you can't play them in real time to see how they look.

Here's one trick I've found - if you're trying to see how the animation looks in real time, you can render it using the "OpenGL Basic" renderer instead of iRay or 3Delight. The OpenGL renderer is VERY fast - it's what's used to do the main window in DS. So each frame renders in seconds, rather than minutes. This lets you render out the animation, play it outside of DS, and decide what needs to be tweaked without taking hours for each iteration.

But khumak is correct - to make things look even vaguely realistic, you need to have "other parts" of the figures move slightly as well. Doesn't have to be much.

To me, the Stimuli animations are nice, but some of them seem a bit "violent" to me. Plus, some of them have been used to death... LOL

 

[khumak]

One thing I've noticed that seems to be common for games that do a lot of animation is for the parts of the characters that would otherwise remain motionless to all just do a slight, gradual turn in 1 direction. I'm not sure what technique they're using to do that but I suspect it might be something I saw in a tutorial video where you can have an object "point at" something.

So you could for instance set a character's eyes, head, forarms/etc to all point at some null object and then just move that null object slightly in 1 direction for each frame and then only manually adjust the position of the part's you really care about. It still looks a bit odd but it does look better than having significant portions of their bodies be completely motionless while the rest of them move.

I'm admittedly still pretty bad at animations. Still trying to figure out a method that would work for me to get them to look smooth. Most of the ones I've tried look fairly jerky.

 

[Rich]

I'm admittedly still pretty bad at animations. Still trying to figure out a method that would work for me to get them to look smooth. Most of the ones I've tried look fairly jerky.

I'm FAR from an expert either. Basically, the Daz Studio animation tool isn't very good - far below what other packages have. But that's because animation is really a secondary "thing" for Daz Studio.

That being said, a couple of ways to help get around "jerky":

1. Where possible, don't put a keyframe in on every Daz frame. Try to spread them out where you can. Daz will smoothly interpolate over intermediate frames. So start by keyframing just the extreme ends of the motion, and then refine. When you look at the Stimuli animations, you see a keyframe on every frame, but I think that's because they were prepared in another package and imported - probably either as BVH or FBX - and the keyframes were generated automatically.
2. When you spread out keyframes, sometimes the body part acceleration/deceleration isn't correct as a result because DS doesn't interpolate the way you'd want it to. The Daz store has a tool called GraphMate that will help - it lets you adjust the shape of the interpolated curves. There is also a tool called KeyMate that lets you do advanced editing of keyframe positions, etc. - much better than the default animation editor DS includes.
3. Consider using another package that does a better job of animation. You could export your figure(s) to Blender, for example, animate them their, and then import just the animation back in. I'm just beginning to experiment with this myself.
4. Practice, practice, practice.... LOL

 

[HandofVecna]

If you mean sex animations, IMO the biggest hurdle to making good ones is the fact that the human eye just seems hypersensitive to movement. So if you have an animation where parts of the figures are moving while other parts are completely motionless with respect to the rest of the figure, it just looks wierd. Unfortunately that means you either end up having to do close ups to reduce the amount of posing you have to do for each frame, or you need to manually change joint positions for just about every major joint in the figure in such a way that it still looks smooth.

The best solution I've found is to use something like one of the animation posing series done by Stimuli (

You must be registered to see the links

) and then you just need to manually tweak minor aspects of each frame like hand/finger position and genitals depending on what angle you're using.

Hey sorry to bother you but I have a question regarding the Stimuli asset. I purchased it, loaded by figs and put them in the poses, moved them onto the bed, adjust, hit animate and they just jump back to the floor. How do I get them to stay on the bed. I tried to parent them to the bed but no go. I really need some help as this is for a project I am working on. I emailed the dev but he's not responded.

 

[khumak]

Hey sorry to bother you but I have a question regarding the Stimuli asset. I purchased it, loaded by figs and put them in the poses, moved them onto the bed, adjust, hit animate and they just jump back to the floor. How do I get them to stay on the bed. I tried to parent them to the bed but no go. I really need some help as this is for a project I am working on. I emailed the dev but he's not responded.

There may be an easier way, but what I do is set up the figures, load the animation, then load the environment and move the environment into position around the actors instead of the other way around. The animation seems to reset each frame back to the 0,0 position.

 

[Porcus Dev]

You can also select two (or more) characters and create a group; and you can move that group wherever you want without affecting the animation (be sure to move it in frame 0).

As for animating the characters with transparent background, yes, it is possible; then render the background only once and put it all together with some video editing program (better than with an image editing program, since with the video editing program you only have to put the background image once throughout the sequence, instead with the image editing program you will have to add the background to each frame you have created, although it can also be done but it takes longer).

If you do it this way (character and background separately), you will have to be careful with the light, because if you remove the scenario the light won't have the same effect on the character (it won't bounce on the walls for example); so the best thing is, first render the background and then render only 1 frame with the character, join them with an image editing program and check if the light is adequate.

Evidently with this system the scene is not so "natural", but it is much faster; I try to render everything together whenever I can because there is a difference in light or shadows, etc; I try to optimize the scene to the maximum and erase the objects that are not seen or interact with the light so that the rendering time of each frame is acceptable (less than 30 minutes if possible ... if it is a very short scene even 1h per frame is acceptable). The character+background system has only been used once with an animation of 250 frames (if I remember correctly) which was much faster with this system, but I try to avoid it.

And if you create your own animations, the "Toogle Pins" utility in "ActivePose" will be your friend

 

[mickydoo]

I cheat, I use the iray interior cam for animations and render the background with it, then render the characters, that way the light is the same as you are using the HDRI, I keep the bed or couch they are on for the shadows of the characters. Due to hardware limitations I only render 10 frames, and reverse it, loop it and speed it up in premier pro. I will upgrade my system in a month or so and I'll change how I do it, but for now its a quick dirty fix.

mgomez0077
I make my animations, using the ActivePose and pins, it's the best way to do it I agree. I get your normal sex poses and use them a lot of times, pin and drag around.

 

[Porcus Dev]

I cheat, I use the iray interior cam for animations and render the background with it, then render the characters, that way the light is the same as you are using the HDRI, I keep the bed or couch they are on for the shadows of the characters. Due to hardware limitations I only render 10 frames, and reverse it, loop it and speed it up in premier pro. I will upgrade my system in a month or so and I'll change how I do it, but for now its a quick dirty fix.

mgomez0077
I make my animations, using the ActivePose and pins, it's the best way to do it I agree. I get your normal sex poses and use them a lot of times, pin and drag around.

I have the iray camera but I haven't tried it yet (it's in the list of many things I have pending, LOL), I think it speeds up a bit the rendering by interpreting/blocking the light better using iray, is that so? or does it have another utility? (Normally I don't use HDRI, I usually use Dome/Scene, Scene Only or Sky-Light).

Erase the whole stage but leave the bed next to the characters for example is very good, because as you say in this way the shadows projected on it make the animation more credible.

I also started with short animations of 10-20 frames, but now I try to create more detailed animations (on average, between 60-150 frames), not only loops, and that there is movement of other parts of the body, not only the genitals, haha (expression, hands, movement of the breasts, etc) ... sometimes it's complicated because DAZ is not specially designed for it but as I do more tests I learn new things.
Upgrading from a 1060 6GB to a 1080Ti 11GB (now I work with 2) has been a great advance for animations (for individual scenes usually with the 1060 I had enough, but moving to the 1080Ti was spectacular, although the animations still took their time, so I have expanded to 2 and I'm ready to continue to expand in the future with a 1300w PSU.... God! Working in this is expensive, hahahaha )

BTW, using "Toggle Pins", the option to activate "Transform Gizmo" is also very useful!

 

[mickydoo]

mgomez0077
I use this HDRI with the camera in an attempt to mimic a room light

You must be registered to see the links

I also made this one myself, just the colour of cream walls and white ceiling with a HDRI exposure cheat in photoshop but it makes it to unrealistic for me.

I only use the HDRI camera for animations, Ive been known to use it for other shots but its too hit and miss unless you use the same 3 or 4 HDRI's I know that work, meaning the render lighting all tends to look the same. In my next release I have a big series of scenes talking to ppl in 5 different locations in a library, I used the HDRI camera for that because it's easier rather than better.

As far as speeding up rendering I don't know, I have have never timed any, I watch them for 5% and see how long it took so I know it's going to work then I watch netflix or fuck off up town.

 

[HandofVecna]

There may be an easier way, but what I do is set up the figures, load the animation, then load the environment and move the environment into position around the actors instead of the other way around. The animation seems to reset each frame back to the 0,0 position.

Thank you! And everyone else. I'll give this a shot.

If it's not too much to ask...

Couple of additional questions if you don't mind when you run animation what are you setting for min update sample, update interval, min sample, max samples and max time?
If any of you use Render Que what are your settings you use? On a one off basis I will set the metrics high and sit here and say ok that looks good, stop render. But when I do a huge series/batch one after another I'm not sure where I should be cutting off at on these.

You guys by far and away have been a great help and I really appreciate it.

 

[Porcus Dev]

Thank you! And everyone else. I'll give this a shot.

If it's not too much to ask...

Couple of additional questions if you don't mind when you run animation what are you setting for min update sample, update interval, min sample, max samples and max time?
If any of you use Render Que what are your settings you use? On a one off basis I will set the metrics high and sit here and say ok that looks good, stop render. But when I do a huge series/batch one after another I'm not sure where I should be cutting off at on these.

You guys by far and away have been a great help and I really appreciate it.

In animations I usually do it as if it were a single image: max samples 5000, max time 0 (infinite), enable rendering quality ON and 95% of percentage.
The rest I leave as it is.
Then you can check how the first frame is rendered; if the time is good I leave it like this, if I see that it takes too long, first I check the scene to see if I can optimize it more (reduce textures, remove objects that are not seen, improve lighting), and if it still takes too long you can reduce the number of iteractions that can vary greatly depending on the scene, sometimes with 500 will look good and other times you will need thousands.

 

[HandofVecna]

In animations I usually do it as if it were a single image: max samples 5000, max time 0 (infinite), enable rendering quality ON and 95% of percentage.
The rest I leave as it is.
Then you can check how the first frame is rendered; if the time is good I leave it like this, if I see that it takes too long, first I check the scene to see if I can optimize it more (reduce textures, remove objects that are not seen, improve lighting), and if it still takes too long you can reduce the number of iteractions that can vary greatly depending on the scene, sometimes with 500 will look good and other times you will need thousands.

Thanks so much! Really great help!

 

[mickydoo]

I do mine with way way less iterations, animations don't need as much as static images as the movement hides flaws. I never use rendering quality enabled for any render. I set the max to 500 most times (using the denoiser), render the image at 3200x1800 and scale it down to 1920x1080. Animations I render at 1920x1080 to start with, I do the back ground first, then render the charcters, this frame has only 200 iterations.I only have a 1050ti card, those two by themselves with just the bed bit under them took 10 minutes or so.

 

[khumak]

I do mine with way way less iterations, animations don't need as much as static images as the movement hides flaws. I never use rendering quality enabled for any render. I set the max to 500 most times (using the denoiser), render the image at 3200x1800 and scale it down to 1920x1080. Animations I render at 1920x1080 to start with, I do the back ground first, then render the charcters, this frame has only 200 iterations.I only have a 1050ti card, those two by themselves with just the bed bit under them took 10 minutes or so.

View attachment 319470

Interesting. I hadn't considered rendering it at a higher res and then downscaling it after the fact. I have the same card you do, but my renders seem to take about twice as long to get to the same point as yours even if I do something similar and hide everything but the couch and the actors. I do have a few extra objects in this one, but I only get about 10 iterations per minute. 500 iterations took me 47 minutes. 500 did seem to be sort of the magic number where almost all of the fireflies go away though. There were some noticeable groups of them until that point, especially if you stopped at 300 or less. Probably wouldn't be that noticeable after downscaling though. This one still has a section with some fireflies for the girl on the right at her neck and down by her left hand, but I can't tell after downscaling.



Do the 3 towels, the skirt, and the panties in my shot make that much difference in the render time? Or maybe the wet skin texture? I tested using the same 3200x1800 resolution that you are using and then downscaled to my usual 1280x720.

I've normally been rendering the entire scene at 1280x720 and it takes about 2 hours to finish with the same 2 characters plus the environment at about 2000-3000 iterations. I organize most of the objects in my scene into groups based on what position they're in relative to the actors so I can easily hide every object that's not visible in the render with about 3 or 4 mouse clicks and that makes a very noticeable difference in render time, sometimes cutting more than 2 hours off my times. The shot below took about 2 hours rendering at 1280x720 with everything included in the shot. This was about 3000 iterations. They haven't oiled themselves up yet though, I don't know how much difference that makes.



I don't hide the walls, floor, or ceiling because I wanted the lighting to be consistent. If I do a shot where I can zoom in and just show the couch plus the actors then it only takes about 20-30 mins or so at that resolution (with everything else hidden other than the walls, floor, and ceiling). I usually set the max iterations and the max time really high and then go watch TV for an hour or two and just stop it when I don't see fireflies anymore. Sometimes that's after 45 mins or something. If there's mirrors involved it might be more like 4 or 5 hours so I try to avoid those.

I could see rendering the background separately saving a bunch of time if you're not changing your camera angle so you can just reuse the background for each shot which is probably the most common case for an animation. I tend to change my camera angles frequently though so I would have to render the background again for every shot most of the time unless it's an animation.

I'm rendering my background for the first shot above now and it's taking a bit longer than the actors plus the couch so I'm guessing 500 iterations will take about 90 minutes. So in my case the finished scene takes about the same amount of time whether I render the background separately or not. I'm definitely curious whether it would be faster to render at a higher res and then downscale after fewer iterations compared to rendering at my target res and just using the image as is when it finishes rendering.

Edit: My background image just finished and took almost exactly 60 mins to render 500 iterations. So I think I actually do save a little bit of time doing it your way. Will have to play around with it a bit more. Here's that first image with the background added, total render time 1 hour 47 mins, which is less time than the one I rendered all in 1 below it and I think it may also be a little clearer.



Edit2:

I was curious so I rerendered the above image at 3200x1800 with the background included. I forgot to come back and check on it so it actually ran to 660 iterations which took a little less than 3 hours before I stopped it. I can't tell a difference in the background, but the lighting is quite a bit different for the 2 girls with the background than it was when rendered separately. It could just be that I don't know how to set the lighting up to be equivalent without the background.

 

[mickydoo]

Have you got the render quality enabled on? I always turn it off, set iterations to max of 500 and use the denoise filter set at 499 (you need daz 4.11 for the denoiser). The rendering images separately only works with the interior cam using a HDRI for the light, if you are using indoor lighting you need the scene for the reflections or they will always look like they do not belong in the scene. The interior cam has it's downsides though, you can get weird reflections on walls, reflections on windows so I tend to only use it for animations. When i do scenes with indoor lighting, I render the background first then spot render the characters, by the time you do it and put it together probably takes the same time as doing it all at once, but you have the background you can use over and over, and the render can concentrate more on the character.