Here are the latest updates to the Crowd Gen Scripts. a full update will come through Daz but that takes a bit longer.
Hopefully fixing all the "nekkid checker being stupid" and also some enhanchements and other bugs fixed.
Sorry for the inconvenience, but this has been such a moving living thing with so many things effecting it from different directions.
OK, I'll try the revised one. It didn't generate any 'nekkid' folks when it couldn't find clothes, must be prudish. LOL.
The original one did, in some weird corner cases, then the Nekkid was added, but was a little to prudish.......
Lowpi Crowd Creator Undead expansion is now out! This expansion adds a brand new lowpoly skeleton figure, zombie skins and clotrhing for Lowpi as well as skeleton and zombie crowd presets.
Yeah, even though Lowpi is rigged similiarly to Genesis 9, it's its own figure, with fewer bones, and therefore cannot take any of Genesis' (any interation) poses.
But, you'll be happy to know that there are more Lowpi poses coming. ;)
I would suggest starting with the base poses that are included and adjusting as needed. Like Fey said, its a very simple figure to cut down both on mesh for unneeded details and things and also to cut down on rigging which doesnt seem like it takes much memoery, but when you have a hundred figures in a scene it adds up. Sicne Lowpi Is really meant to be used from a distance ( as in a block or more away) poses don't need to be perfect.
I have already created some wear kits and pose kits for myself. I have also saved several crowd setups from the crowd generation script.Now I have the Undead expansion. Do I have to go back and edit any wear kit or pose kit I have already created, to add the zombies in as being compatible with the clothing or poses I have created? And then do I have to recreate all sets that use those wear kits and pose kits? And then do I have to recreate all crowd setups I have saved, that use those sets? If I have to do this every time a new expansion is released,. this will become an overwhelming amount of work.
I have already created some wear kits and pose kits for myself. I have also saved several crowd setups from the crowd generation script.Now I have the Undead expansion. Do I have to go back and edit any wear kit or pose kit I have already created, to add the zombies in as being compatible with the clothing or poses I have created? And then do I have to recreate all sets that use those wear kits and pose kits? And then do I have to recreate all crowd setups I have saved, that use those sets? If I have to do this every time a new expansion is released,. this will become an overwhelming amount of work.
The basic thought is that you have multiple WearKits, PoseKits and PropKits that you create LowPi Kits from. Now when Undeads was released, if your clothing sets works with Undeads, the best practice is to load your WearKit Save it out with a new name, replace the Compatibilities to be Zombies instead and save it. Then if you want to use a PoseKit, clone, rename and change the Compatibilties. The reason for this when it comes for Zombies is that Zombies do have their own compkeys (which defines which morphs to randomize when generating a Zombie.
But this gives me an idea for a Set Builder, that maybe should have a clone function where you replace/add compatibilities. I have started to do come Compatibilties fixes in Set Builder, fix issues where you have a "bad" Compatibility set on something so you can clear that out, and also clear out for example poses for child of you have a Set that does not now any WearKits for child, but have poses for Child, which can cause issues. I should probablyu add a "Retarget" option to say I want these Compatibilities Added and these removed, will look into that.
In the wear kit section I see a Full Body option in the Wearable Type. If I select Full Body, (like for my one piece combat uniform), do I have to set "incompatible" for every Top and Skirt and Pants item, or does the script automatically understand that incompatibility?
In the wear kit section I see a Full Body option in the Wearable Type. If I select Full Body, (like for my one piece combat uniform), do I have to set "incompatible" for every Top and Skirt and Pants item, or does the script automatically understand that incompatibility?
You should create one WearKit using the Full Body, which is great for full wearable sets, like the Skeleton, or if you save out a set of clothes as a Wearable Preset, and you can then add hais to that Weakit.
Then you should create a second WearKit of you have multiple single loadable items,
The WearKit options are just "tags" defined in wearkeys, anf the logic really picks one from each category that is defined in that WearKit, that's why you should build multiple small WearKits instead of one huge, as you can then control the presence on a per WearKit basis, I want 1/3 (5) wearing this full set and 2/3 (10) wearing clothes from this set of more random parts, and it's much easier to setup. Also, if you use a FullBody based on a Wearable Preset, you need to use a Hierachical HMAT material to define "the whole set" as it will load the items in one load, a way to get "color matching top and pants for example" as there is no way to identify each item and to get hold of it and then apply separate mats for each item (as they are loaded in one load operation by DS), and you cannot define that anyway in the WearKit Builder, but you can add a HMAT as a material setting.
Thank you for the wear kit tips. I have a hat. I tagged it as hair. Does that seem correct? It doesn't fit over hair. It is used instead of hair. I couldn't see any other tag that would make sense for a hat.
Thank you for the wear kit tips. I have a hat. I tagged it as hair. Does that seem correct? It doesn't fit over hair. It is used instead of hair. I couldn't see any other tag that would make sense for a hat.
That is exactly how is should be done as you do not need "hair" at that distance if you have a hat.
You could use a geoshell and a painted hair together with the hat (as a wearable preset ) if you want to go overboard.
The LowPi crowd generation system is quite complex and I know people will figure out clever ways of doing things that we never thought of.
I stripped down my military wear kit to the minimum, based on Totte's recommendations. I was finally able to resolve my naked people problem by removing my wearable preset with the uniform and hat, and just left two separate items, the uniform and the hat. I also added some hair to the wear kit.
This is a race to reach the last helicopter out of the jungle. I made a crowd of 50 LowPi, but some are not visible behind the large helicopter. There is an attraction cone with strength 3 where the helicopter is sitting.
This looks incredible. I might buy something in the Daz store for the first time in a while.
I was wondering if an expansion could possibly add some models with a bit more detail for ones closer to the camera? If lowpi characters are looking at the camera while being anywhere near the camera they kind of stand out.
But instead of trying to make new lowpi models, which I understand is already quite complicated, it could be done by utilizing the clothing script. Why not just 'wear' a more detailed head prop as clothing? That way, it could still fit within the existing script without changing anything. "Head Props for Lowpi", or whatever. They might need to hide the head to ensure it doesn't clip through, but it sounds doable. Perhaps a new script product that lets users behead their favorite Genesis characters and turn them into wearable props for lowpi? As a bonus, it would give users the ability to alter textures that have familiar Genesis UVs. Too many would still present an issue for DS, but again, these would be limited to the ones more visible to the camera. Maybe one could add some replacement hands as well, especially for zombies which are often shown moving with their hands open towards the humans. So you could have a group of zombies closer to the human (and camera) that have their heads and hands replaced with wearable props, with a couple dozen more standard zombie lowpis in the distance behind them. That would be epic.
At any rate, being able to generate crowds for Daz like this is a pretty big deal. The really cool part isn't just being able to generate a crowd, it is being able to automate the poses to a degree for them. You have a product that you could be creating expansions to for years to come.
This is also an important product because it is something that Stable Diffusion simply cannot do with any consistency. Try making a crowd in any image generator, you are guaranteed to have at least a few broken faces, extra limbs, and people melting together. Daz needs to push products like this which demonstrate its strengths over image generators.
Awesome work, barbult. You are inspiring me to attempt making kits.
I love the idea of using a geoshell for a wearable, Totte.
I wanted to mention there is one pose that is in the incuded pose kits that just doesn't work for standing figures - that is the lounging top. It looks dumb, even from a distance.
I noticed today that all my LowPi in the helicopter image are posed basically the same way, with the right leg forward, running. I know that the attention marker strength modifies the pose to make them run if the attention value is strong, but I thought it would just exaggerate the existing pose. Does it completely change the pose to a right leg forward running pose? I'll have to turn down the attention strength and see if they all started with right leg forward or whether they got changed. It does look strange to see them all running in the same pose.
I wanted to mention there is one pose that is in the incuded pose kits that just doesn't work for standing figures - that is the lounging top. It looks dumb, even from a distance.
When I made the base poses for Lowpi, we didn't yet have the city props, and Totte was still chunking away at buildinig the Crowd Builder scripts. I threw the lounging pose in because I wanted to cover the basics - standing, sitting, laying down, walking, and running. (Actually, I have a tendency to build far more poses than is required for any given situation, so Lyrra gave me a strict "do five poses to cover the basics" ) For the base Crowd Builder stuff, there was no need to really use the Lounging pose in the Posekits It's just there for customers who don't buy the bundle or buy the Proscripts. If we ever get around to building an Orgy expansion, I'm sure the Lounging pose will see some love (as it were. Lol.)
As for the Lounging Top, I think that just was thrown into the mix (if I'm remembering correctly...it's been a while) so as to cover different ranges of movements for the Posekits. That's also the reason why there are tops and bottom presets in the poses, to easily use them in the Posekits - now and in the future. The nice thing is that you can replace it, or any other weird pose that gets generated by selecting that figure (assuming it's not an instance) and either tweaking the bone dials in the parameters tab, or just select another top from the poses that is more in keeping with what you had intended for your scene.
I noticed today that all my LowPi in the helicopter image are posed basically the same way, with the right leg forward, running. I know that the attention marker strength modifies the pose to make them run if the attention value is strong, but I thought it would just exaggerate the existing pose. Does it completely change the pose to a right leg forward running pose? I'll have to turn down the attention strength and see if they all started with right leg forward or whether they got changed. It does look strange to see them all running in the same pose.
It sounds to me like the Crowd Generator is pulling from the right hand version of the pose more than the left hand version. I'm not sure why that is. It's a question for Totte, I'm afraid.
I noticed today that all my LowPi in the helicopter image are posed basically the same way, with the right leg forward, running. I know that the attention marker strength modifies the pose to make them run if the attention value is strong, but I thought it would just exaggerate the existing pose. Does it completely change the pose to a right leg forward running pose? I'll have to turn down the attention strength and see if they all started with right leg forward or whether they got changed. It does look strange to see them all running in the same pose.
It sounds to me like the Crowd Generator is pulling from the right hand version of the pose more than the left hand version. I'm not sure why that is. It's a question for Totte, I'm afraid.
We'll wait for his take on it, then.
This is what my crowd looks like with attention = 0, for reference.
When I generate a skeleton crowd, why is each skeleton parented to an invisible LowPi character?
Because the skeleton is a wearable "clothing" (that was Lyrra's trick to make a skeleton based on LowPi, which also shows the possibilities you can use to create very different "clothings"
When I generate a skeleton crowd, why is each skeleton parented to an invisible LowPi character?
Because the skeleton is a wearable "clothing" (that was Lyrra's trick to make a skeleton based on LowPi, which also shows the possibilities you can use to create very different "clothings"
The skeleton wearable has rigging of its own and can take the LowPi poses. Does it have to be parented to a LowPi to get the scripts to work in it. Is that the reason it is "clothing" and not a figure of its own?
I noticed today that all my LowPi in the helicopter image are posed basically the same way, with the right leg forward, running. I know that the attention marker strength modifies the pose to make them run if the attention value is strong, but I thought it would just exaggerate the existing pose. Does it completely change the pose to a right leg forward running pose? I'll have to turn down the attention strength and see if they all started with right leg forward or whether they got changed. It does look strange to see them all running in the same pose.
How attention works:
Each pose has a speed (defined from 0 (still) to 5 (dashing)
The attention markers effect 0 -> +/-5 will call for what internally is called "a speed pose". When Crowd generator is preparing the data to send to the popengine, it walks through all poses in all posekits and creates a separate bucket called speedposes indexed from 1 - > 5 (as zero is not moving), then if an attentionmarker comes into play, it will try to find a pose matching the speed the attention pole asks for, and if it cannot find an exct match, it looks for all speedposes and randomizes among them.
When I generate a skeleton crowd, why is each skeleton parented to an invisible LowPi character?
Because the skeleton is a wearable "clothing" (that was Lyrra's trick to make a skeleton based on LowPi, which also shows the possibilities you can use to create very different "clothings"
The skeleton wearable has rigging of its own and can take the LowPi poses. Does it have to be parented to a LowPi to get the scripts to work in it. Is that the reason it is "clothing" and not a figure of its own?
Yes, it needs to be worn by LowPi for the underlaying functionality to work.
I noticed today that all my LowPi in the helicopter image are posed basically the same way, with the right leg forward, running. I know that the attention marker strength modifies the pose to make them run if the attention value is strong, but I thought it would just exaggerate the existing pose. Does it completely change the pose to a right leg forward running pose? I'll have to turn down the attention strength and see if they all started with right leg forward or whether they got changed. It does look strange to see them all running in the same pose.
It sounds to me like the Crowd Generator is pulling from the right hand version of the pose more than the left hand version. I'm not sure why that is. It's a question for Totte, I'm afraid.
We'll wait for his take on it, then.
This is what my crowd looks like with attention = 0, for reference.
Need to check the PoseKit but I guess it might have both versions setup as upper poses for that lower pose and randomly picks one, or the posekit might have issues, been a real pain to have to update them after rig updates etc. PoseKits can be easily tweaked in PoseKit Builder and saved out to use them with your own WearKits, and also, the jitter settings can me refined a lot, depnding on your likings...
Speaking of skeletons, why does it have a displacement map? I don't understand how a gray-ish displacement map, with a strength of 0.01, default min and max displacement, on a base resolution character, rendered in Iray can possibly have any impact. I can't see any difference, even in very closeup renders. What am I missing?
Comments
The original one did, in some weird corner cases, then the Nekkid was added, but was a little to prudish.......
Lowpi Crowd Creator Undead expansion is now out! This expansion adds a brand new lowpoly skeleton figure, zombie skins and clotrhing for Lowpi as well as skeleton and zombie crowd presets.
https://www.daz3d.com/lowpoly-crowd-creator-expansion-undead
I converted a G9 combat outfit to fit LowPi and added it to a wear kit. I added a skeletons set. Here we are.
Great images @barbult!
Can any other poses from daz figures be used on these figures? I can't seem to manually pose them.
From earlier in this thread: "The rig is very similar to Genesis 9, but fingers and toe poses might get wonky, depending the specific pose used"
Yeah, even though Lowpi is rigged similiarly to Genesis 9, it's its own figure, with fewer bones, and therefore cannot take any of Genesis' (any interation) poses.
But, you'll be happy to know that there are more Lowpi poses coming. ;)
I would suggest starting with the base poses that are included and adjusting as needed. Like Fey said, its a very simple figure to cut down both on mesh for unneeded details and things and also to cut down on rigging which doesnt seem like it takes much memoery, but when you have a hundred figures in a scene it adds up. Sicne Lowpi Is really meant to be used from a distance ( as in a block or more away) poses don't need to be perfect.
LM
I have already created some wear kits and pose kits for myself. I have also saved several crowd setups from the crowd generation script.Now I have the Undead expansion. Do I have to go back and edit any wear kit or pose kit I have already created, to add the zombies in as being compatible with the clothing or poses I have created? And then do I have to recreate all sets that use those wear kits and pose kits? And then do I have to recreate all crowd setups I have saved, that use those sets? If I have to do this every time a new expansion is released,. this will become an overwhelming amount of work.
The basic thought is that you have multiple WearKits, PoseKits and PropKits that you create LowPi Kits from. Now when Undeads was released, if your clothing sets works with Undeads, the best practice is to load your WearKit Save it out with a new name, replace the Compatibilities to be Zombies instead and save it. Then if you want to use a PoseKit, clone, rename and change the Compatibilties. The reason for this when it comes for Zombies is that Zombies do have their own compkeys (which defines which morphs to randomize when generating a Zombie.
But this gives me an idea for a Set Builder, that maybe should have a clone function where you replace/add compatibilities. I have started to do come Compatibilties fixes in Set Builder, fix issues where you have a "bad" Compatibility set on something so you can clear that out, and also clear out for example poses for child of you have a Set that does not now any WearKits for child, but have poses for Child, which can cause issues. I should probablyu add a "Retarget" option to say I want these Compatibilities Added and these removed, will look into that.
In the wear kit section I see a Full Body option in the Wearable Type. If I select Full Body, (like for my one piece combat uniform), do I have to set "incompatible" for every Top and Skirt and Pants item, or does the script automatically understand that incompatibility?
You should create one WearKit using the Full Body, which is great for full wearable sets, like the Skeleton, or if you save out a set of clothes as a Wearable Preset, and you can then add hais to that Weakit.
Then you should create a second WearKit of you have multiple single loadable items,
The WearKit options are just "tags" defined in wearkeys, anf the logic really picks one from each category that is defined in that WearKit, that's why you should build multiple small WearKits instead of one huge, as you can then control the presence on a per WearKit basis, I want 1/3 (5) wearing this full set and 2/3 (10) wearing clothes from this set of more random parts, and it's much easier to setup. Also, if you use a FullBody based on a Wearable Preset, you need to use a Hierachical HMAT material to define "the whole set" as it will load the items in one load, a way to get "color matching top and pants for example" as there is no way to identify each item and to get hold of it and then apply separate mats for each item (as they are loaded in one load operation by DS), and you cannot define that anyway in the WearKit Builder, but you can add a HMAT as a material setting.
Thank you for the wear kit tips. I have a hat. I tagged it as hair. Does that seem correct? It doesn't fit over hair. It is used instead of hair. I couldn't see any other tag that would make sense for a hat.
That is exactly how is should be done as you do not need "hair" at that distance if you have a hat.
You could use a geoshell and a painted hair together with the hat (as a wearable preset ) if you want to go overboard.
The LowPi crowd generation system is quite complex and I know people will figure out clever ways of doing things that we never thought of.
I stripped down my military wear kit to the minimum, based on Totte's recommendations. I was finally able to resolve my naked people problem by removing my wearable preset with the uniform and hat, and just left two separate items, the uniform and the hat. I also added some hair to the wear kit.
This is a race to reach the last helicopter out of the jungle. I made a crowd of 50 LowPi, but some are not visible behind the large helicopter. There is an attraction cone with strength 3 where the helicopter is sitting.
Another masterpiece usage of LowPi!
Very, very nice, @barbult
I think to not showing the faces of LowPi is a key to nice image.
This looks incredible. I might buy something in the Daz store for the first time in a while.
I was wondering if an expansion could possibly add some models with a bit more detail for ones closer to the camera? If lowpi characters are looking at the camera while being anywhere near the camera they kind of stand out.
But instead of trying to make new lowpi models, which I understand is already quite complicated, it could be done by utilizing the clothing script. Why not just 'wear' a more detailed head prop as clothing? That way, it could still fit within the existing script without changing anything. "Head Props for Lowpi", or whatever. They might need to hide the head to ensure it doesn't clip through, but it sounds doable. Perhaps a new script product that lets users behead their favorite Genesis characters and turn them into wearable props for lowpi? As a bonus, it would give users the ability to alter textures that have familiar Genesis UVs. Too many would still present an issue for DS, but again, these would be limited to the ones more visible to the camera. Maybe one could add some replacement hands as well, especially for zombies which are often shown moving with their hands open towards the humans. So you could have a group of zombies closer to the human (and camera) that have their heads and hands replaced with wearable props, with a couple dozen more standard zombie lowpis in the distance behind them. That would be epic.
At any rate, being able to generate crowds for Daz like this is a pretty big deal. The really cool part isn't just being able to generate a crowd, it is being able to automate the poses to a degree for them. You have a product that you could be creating expansions to for years to come.
This is also an important product because it is something that Stable Diffusion simply cannot do with any consistency. Try making a crowd in any image generator, you are guaranteed to have at least a few broken faces, extra limbs, and people melting together. Daz needs to push products like this which demonstrate its strengths over image generators.
Awesome work, barbult. You are inspiring me to attempt making kits.
I love the idea of using a geoshell for a wearable, Totte.
I wanted to mention there is one pose that is in the incuded pose kits that just doesn't work for standing figures - that is the lounging top. It looks dumb, even from a distance.
I noticed today that all my LowPi in the helicopter image are posed basically the same way, with the right leg forward, running. I know that the attention marker strength modifies the pose to make them run if the attention value is strong, but I thought it would just exaggerate the existing pose. Does it completely change the pose to a right leg forward running pose? I'll have to turn down the attention strength and see if they all started with right leg forward or whether they got changed. It does look strange to see them all running in the same pose.
When I made the base poses for Lowpi, we didn't yet have the city props, and Totte was still chunking away at buildinig the Crowd Builder scripts. I threw the lounging pose in because I wanted to cover the basics - standing, sitting, laying down, walking, and running. (Actually, I have a tendency to build far more poses than is required for any given situation, so Lyrra gave me a strict "do five poses to cover the basics" ) For the base Crowd Builder stuff, there was no need to really use the Lounging pose in the Posekits It's just there for customers who don't buy the bundle or buy the Proscripts. If we ever get around to building an Orgy expansion, I'm sure the Lounging pose will see some love (as it were. Lol.)
As for the Lounging Top, I think that just was thrown into the mix (if I'm remembering correctly...it's been a while) so as to cover different ranges of movements for the Posekits. That's also the reason why there are tops and bottom presets in the poses, to easily use them in the Posekits - now and in the future. The nice thing is that you can replace it, or any other weird pose that gets generated by selecting that figure (assuming it's not an instance) and either tweaking the bone dials in the parameters tab, or just select another top from the poses that is more in keeping with what you had intended for your scene.
It sounds to me like the Crowd Generator is pulling from the right hand version of the pose more than the left hand version. I'm not sure why that is. It's a question for Totte, I'm afraid.
We'll wait for his take on it, then.
This is what my crowd looks like with attention = 0, for reference.
When I generate a skeleton crowd, why is each skeleton parented to an invisible LowPi character?
Because the skeleton is a wearable "clothing" (that was Lyrra's trick to make a skeleton based on LowPi, which also shows the possibilities you can use to create very different "clothings"
The skeleton wearable has rigging of its own and can take the LowPi poses. Does it have to be parented to a LowPi to get the scripts to work in it. Is that the reason it is "clothing" and not a figure of its own?
How attention works:
Each pose has a speed (defined from 0 (still) to 5 (dashing)
The attention markers effect 0 -> +/-5 will call for what internally is called "a speed pose". When Crowd generator is preparing the data to send to the popengine, it walks through all poses in all posekits and creates a separate bucket called speedposes indexed from 1 - > 5 (as zero is not moving), then if an attentionmarker comes into play, it will try to find a pose matching the speed the attention pole asks for, and if it cannot find an exct match, it looks for all speedposes and randomizes among them.
Yes, it needs to be worn by LowPi for the underlaying functionality to work.
Need to check the PoseKit but I guess it might have both versions setup as upper poses for that lower pose and randomly picks one, or the posekit might have issues, been a real pain to have to update them after rig updates etc. PoseKits can be easily tweaked in PoseKit Builder and saved out to use them with your own WearKits, and also, the jitter settings can me refined a lot, depnding on your likings...
But will check the FF Moving PoseKit soon.
Speaking of skeletons, why does it have a displacement map? I don't understand how a gray-ish displacement map, with a strength of 0.01, default min and max displacement, on a base resolution character, rendered in Iray can possibly have any impact. I can't see any difference, even in very closeup renders. What am I missing?