Project Bento Feedback Thread

[kayla Mekanic]

Why do we have to keep delaying the full release? Bento is pretty much good to go. The only reason it's being held back is because of this animation/sliders issue. Unfortunately we cant have both at the same time. I think creators (me included) could just choose whether to rig to the sliders or not. Plus Bento is mainly about new bones for rigging and animation, not slider customization...

[wherorangi]

thanks Medhue (:

i think you have shown enough to satisfy any reasonable person, and is really good that you have

i really like what you have been able to do and show, and that you took the time to do so. And bc you have then I think also that the tradeoffs are easy understandable. So thankyou for all of this (: 

and if Bento comes to SL proper, pretty much like this, then I will personally be pretty ok and happy about it

 

[Cathy Foil]

Hi Kayla,
The reason for the delay for the past few weeks are that we are hard at work doing our best to make the new face bones work well with the sliders and for animators.  Once released we won't be able to make any more changes to them since it will break content so we are doing our best to try and find the best possible balance between animation and customization with the Appearance Sliders.

We been mainly doing two things to achieve the best balance possible over the last couple weeks.  One moving some of the bones, mainly the lip bones, so that they are positioned so that rotation animations will work as well as they can on a human mesh head seeing that rotation animations work well for almost all slider settings.

The second thing we been working on is the Avatar Lad code or avatar_lad.xml file.  This file controls how the Appearance Editor sliders work.  Matrice, Vir and I have gone through the code switching as many of the new Bento bones to use scaling instead of translations or what in the LAD is called "offset".  This gives us great benefit because scales do not interfere with animations using translations.  It also has the benefit that non human avatars will be more affected by Appearance Sliders so they will enjoy customization more like the human avatars have.

We are just ironing out the last few glitches and hopefully Bento will be finalized and put in the Release Candidate Viewer in the next few days.

I say there has been a minimum improvement of 30% to Bento in the last two weeks or so and I believe the slight extra wait will have been worth it.

[Pirschjaeger Fassbinder]

Speaking of sliders and non-human heads, I have been experimenting with the sliders with my dragon head on the latest test build. Some of the sliders work fine and others would create weird distortions such as parts of the mesh poking through other parts of the mesh. Here is just one example:

I was finally able to make the head blink using the eyelid bones. So the eyelids were somewhat adjustable with eye size. I could not make the eyes bigger than 50% without either the eyelids not being able to close or them becoming larger than the eye socket itself. However, I was able to make them as small as 24% with some adjustments in the “Eye Opening” slider. I also got varied results with the “Eye Spacing” and “Eye Depth” sliders.

Please note: I might still need to play with the weights.

Of course, I do not expect the sliders to totally work with non-human heads. However, it still allows for some customization. If I manage to sell it, I think I will at least recommend that the original shape be copied first before editing with the sliders.

[MicheloudIgnatio]

Now that project Bento is about to be finalised, are LLs going to take over a new project regarding the compatibility between all those mesh bodies and heads and their clothing and their appliers?

If yes, when?

[Anne Cloud]

I am experiencing some inconsistency with relation to z-offset in animations between the current most recent Bento build (317134) and Vir's currentTest build (317405). 

This first graphic shows the expected animation, which can be seen in the Bento build:

https://i.gyazo.com/41c6a9c373f3b6f25f3f13d8dc346acf.gif

This following graphic shows how the same animation looks seen via the Test build:

https://i.gyazo.com/57f39d5965a90f0f0bd07b24f1e2ff7a.gif

 I am also noticing that, whereas previously the camera would hover at a consistent height above the ground behind the avatar, the camera now bounces up and down *with* the avatar based on the z-offset of the animation. 

Are these behaviours expected? They are subtle but different and in the latter case, may affect general user usability (possibly causing nausea issues).

 

[Anne Cloud]

Even just standing still normally, not cammed in on anything, I notice marked change:

(Edited to fix link: https://i.gyazo.com/263e52ec8b1ead6591067862f4fcde66.gif )

[Cathy Foil]

---

MicheloudIgnatio wrote:

Now that project Bento is about to be finalised, are LLs going to take over a new project regarding the compatibility between all those mesh bodies and heads and their clothing and their appliers?

If yes, when?

---

I been involved in the Bento group since residents were invited to join group under the NDA.  As far as I know LL hasn't mentioned any new project with those goals in mind.

A number of times over the last year some ideas were kicked around about how some code could be written to eliminate the need for appliers, alpha huds and onion skinned avatars.  It has huge potential to reduce lag and make custom mesh avatars and mesh heads easier to use.

If you are talking about some sort of system to make mesh clothing compatible between different mesh bodies which have different shapes and weighting I doubt such a system would be possible.

Theoretically if all mesh clothes were made to fit the LL default avatar and mesh body creators submitted a Morph that changed their mesh body shape into the shape of the LL default avatar a wrap deformer could reshape the clothes to fit the custom mesh body shape.

But then you have the second problem that each custom mesh body has its own unique weights.  So while the clothes may fit while the avatar is standing still as soon as it moves the clothes may not move in the same way or change shape in the same way when the Appearance Editor sliders are changed if the mesh clothes and mesh avatar are both using Fitted Mesh.

Theoretically you could automatically replace the weights of the mesh clothing by copying the weights from the custom avatar mesh.  This would only really work for mesh shirts and pants.  Mesh skirts and dresses the vertex between the legs copying weights don't work well at all.

 

Also theoretically you could have mesh clothing that are not rigged at all and use a wrap deformer to move the clothes with the avatar mesh.  Again this really only works with shirts and pants.  It is also super intensive processing and would slow your computer down to a crawl.  I know I tried it in Maya.  Movement become super jerky and down to about 2 or 3 frames per second and I have a good computer.

I hope the ideas put forth that would eliminate the need for appliers and onion skinned avatar gets adopted into a project and I get an invitation to join the group but nothing official has been said or implied that it might.

[Polymath Snuggler]

Oh wow that is more bizarre than my original understanding of the issue.  I'll see if I can reproduce it 

[Anne Cloud]

Additionally, the pre-jump stick issue continues to be a problem for animation overrides - particularly in cases where AOs are essential towards the believability of non-standard mesh rigs.

(some additional distortion between upper body and pelvis appears to be present in that video, but I have yet to diagnose what is causing the issue.)

[MicheloudIgnatio]

Thanks for the info Cathy

 

So, from what I understand, there are a few things that can be done regararding the compatibility of the mesh body parts (head, body, hands, feet).

I think this is the most important thing to become compatible. Like for example when I wear my catwa head with my Slink body, and I want to change to my niramyth body, I would not have to put on the hud of the head in order to push the button in it that will make the head compatible with niramyth. Or when I wear my Slink hands with my Slink body and I want to move to the system body, I will not have to put on the hud of the hands to press that button that fix the wrists to the system body.

The best thing would be that all this compatibility would happen automaticaly, without any action from the user. And if this is not possible, then at least we should have a panel build in the viewer, that will give these options altogether, so we will not have to attach and dettach a lot of those huds as we do now. 

The same with the various skin appliers. One applier that will cover all the curently used mesh body parts, without the need to attach and detach one hud for the head, another for the body and another one for hands and feet. And again here, if it cannot happen automaticaly, then at least through a build in the viewer panel.

Now regarding clothes as you made me realize there are some constraints, so I will just lIve this one out. But at least the current system with the auto hide script could become a standard, so we will not have to use huds all the time to alpha our body parts from outfit to outfit.

 

[Television]

This is something I've actually given quite a bit of thought to myself. As far as mesh clothes -- the best you can get is fitted mesh, unfortunately.

But as far as appliers go, I've always thought about how much easier it would be if there was a way to transmit avatar bake textures to a mesh attachment. When you wear a default LL clothing layer such as a skin, jacket, pants or tattoo layer, the server will take all of these seperate textures and combine them into one single texture in order to apply it to a single mesh (the default LL body.)

Mesh bodies attempt to replicate this effect the best they can by making multiple copies of the body to wear over each other so people can apply different clothing textures. Not only is it impossible to write a script to make your own "bake" for mesh bodies, but having multiple clothing/tattoo layers over the mesh body itself will often double, triple, or quadrouple the render weight since the clothing/tattoo layers need to have the exact same polys and rigging as the base body to not have z-indexing issues or clipping.

If LL were to allow the transfer of an avatar bake onto an object you would completely eleminate the need for appliers, extra clothing/tattoo layers, etc, making the rendering weight of mesh bodies far more pracitcal and allowing the use of the much cleaner and optimized clothing layer system second life has already had as a default.

My understanding of past concerns about this is that people would assume it would make textures easier to steal, but the reality of it is -- if someone wants to steal textures, it's not that hard. And they certainly wouldn't steal the avatar bake because a baked, compressed texture is completely useless if someone is ripping textures to resell or use. So opening up the UUID of avatar bakes for mesh bodies to utilize would not make the texture thieving situation any better or worse, but it Will allow a much more framerate and loadtime friendly second life.

Perhaps a consideration for future additions.

[Medhue Simoni]

Hmmm, it makes me wonder if the internal breathing animations are affecting your avatar. Sometimes, I do this on purpose. I'd be interested in seeing that animation in Blender.

[Medhue Simoni]

I just want to talk briefly about a comment I made at the last meeting. I didn't want to harp on it more during the meeting, but I'll expand on it here. I talked about how much work it is to keep redoing avatars for all these skeletal changes. Yes, it is fairly simple to convert the skeletons, but that is only 1 small part of updating an avatar. After that, weights need to be adjusted. After that, ALL or most of the animations need to be updated. For my wolf and Coyot avatars, the face alone has more than 30 animations, and then I still have to update all the animations in the AO, and the hud animations. This, ALL, just to test things. And while doing all this, trying to figure out or deal with problems. Oh, and let's not forget any clothing that gets made for these avatars.

My Coyot avatar is now on version 10, and my wolf is on version 9. Many of the other avatars I've made and tested, I just stopped at 1 or 2 versions, strictly for testing specific things. So, my point is, updating an avatar is FAR more complicated than simply updating the rig, as a custom avatar is nothing without the animations.

 

[Medhue Simoni]

Outside of appliers, which LL has talked about, it's not possible to make all the bodies and heads compatible from LL's side. LL has nothing at all to do with these mesh bodies, so how exactly would they make them compatible? If the mesh bodies and parts are to be compatible in any way, it needs to be done by the creators.

I will mention that I have seen many of the mesh bodies in Blender, and some of the weights on these bodies are actually worse than the default weights. IMHO, if things are going to be compatible, the creators of these bodies need to adopt 1 set of weights to go off of, and not every single body using different weights. Take note, I sell weights for fitted mesh avatars. lol

[Vir Linden]

We have a build in QA now containing what is planned to be the final set of skeleton/slider changes. When it releases as a project viewer update, this will affect the appearance of existing content and will likely require some rework of some models. We are sorry that this will inconvenience many people; the decision, based in large part on resident feedback, was that the changes were beneficial enough to offset the negatives, but that does not mean the negatives aren't there, or should be discounted. The positives are substantial, including a much better jaw configuration for animation and shaping, better support for combining shape controls and animations, better positioning of several face bones, a more logical structure for the skeleton, and many improvements to individual sliders.

Going forward, we hope to make no further skeleton/slider changes, and to ultimately go to release candidate and release with these settings. If necessary, we will make changes only to fix major problems, and only the smallest possible changes, in order to affect existing content as little as possible. Since the skeleton and slider configuration involves thousands of largely hand-tuned settings, it is entirely possible that in spite of everyone's best efforts over many months, there may be imperfections remaining in the final configuration that are not major problems, and that we will not be able to change.

When the build goes out, we will post again here, and will update the information about available test models. Since there's a new bone added, we will also need to update the server configuration on aditi and agni to allow uploads that reference that bone.

Thanks to everyone who has contributed to the development of the updated skeleton over the course of the project!

[Tornleaf]

Honestly this is the reason I've refrained from pouring too much work into my bento avatars. I'm too fearful (ie lazy) to spend hours on something that could change, even at this stage - so kudos to those of you who have tested and tested again!

[Anne Cloud]

Here is how that standing animation looks in Blender. No changes have been made to it since the animation was uploaded. It is priority 4. 

https://i.gyazo.com/c596455e31170aac699363cc352663b2.gif

 

[Medhue Simoni]

---

Aki Shichiroji wrote:

Here is how that standing animation looks in Blender. No changes have been made to it since the animation was uploaded. It is priority 4. 

https://i.gyazo.com/c596455e31170aac699363cc352663b2.gif

 

---

What's interesting to me, is where your COG is. It makes me wonder if that COG placement is your whole issue, especially with hover height. Have you ever tried adjusting your body shape? I would think it would have some influence on hover height, possibly because of the COG placement.

[MicheloudIgnatio]

Thanks for the information Medhue.

Unfortunately I am just an end user without any technical knowledge, so I can't suggest any solution.

But the problem unfortunately is there, thats why maybe LL could take the initiative to give some motive to the creators to update their meshes with compatibility in mind. For example LL could intoduce a build in panel for the management of the meshes, but only meshes updated with compatibility in mind could work through it. Thats an idea just from the top of my head, which cannot stand in to much critisism. I just want to stress the importance of the issue, and hope that the LL will take the initiative to bring the creators together in solving this problem. 

And of course this will need to be taken up as a new project and be thouhrouly discussed 

[Anne Cloud]

I should note the COG isn't any higher or lower than normal - it's just that some bones are much higher than they would normally be with an average SL human. 

With that said, it works properly with the most current Bento viewer. I don't know whether this problem is a side-effect of the adjustments Vir made regarding the hover height issue Teager reported or if it is completely separate, but given that both issues involve vertical positioning and it only showed up most recently in the Test viewer, I have a strong feeling that they are related.

I've written up a JIRA for this here: https://jira.secondlife.com/browse/BUG-20169

[Medhue Simoni]

I would imagine that compatibility between all these options is pretty complex, and crazy for the end users. I'm a guy, with hardly any interest in trying to match up parts, so I can only imagine. I have watched some videos of users reviewing many of these options, as I figured these videos would give a good incite into those markets. Plus, I'v rigged countless clothing items for many of the clothing designers that make clothing for these bodies. So, although I can understand the technical side of it all, I have limited understanding of the items in the grand view.

I will say tho, because I believe in market forces and I sell products myself, that the end users have much more power than most can imagine. The problem, to me, is getting the consumers to have a united voice. So, if the end users are united in wanting a particular feature, like compatibility, and express it loudly enough, the creators will respond. Hmmmm, kind of like LL did, lol.

[Cathy Foil]

You hit the nail right on the head.  That is pretty much the same exact system we discussed a few times during a few Bento meetings.  Mind you the meetings were not completely about this topic.  More like 4 or 5 minutes of two or three meetings.

The different layers of textures the LL system default avatar wears are already all baked down to one texture that is then applied to the default avatar already including the alpha layers.

So to make custom avatar mesh work some code needs to be added to take that baked texture or flattened texture and automatically apply it to the custom avatar mesh be it a mesh body or body part or even mesh clothes.

We also need a way in the mesh upload window to designate that the mesh being upload is a mesh that is to be automatically textured with the baked texture your avatar is currently wearing.

In the beginning only mesh that had the same UV layout as the LL default avatar mesh would work.  After the initial release of the system a follow up project might happen that could project that flatten image onto any UV layout allowing for custom UV layouts.

The only remaining hurdle is how to hide your LL default avatar mesh.  Currently we hide the mesh by wearing an alpha that covers the entire body but that means the flatten or baked texture is completely invisible.  Most likely LL could add a button in the viewer so it is not rendered or what I like better is a new Alpha Layer that is not baked to the texture that is applied to the custom avatar mesh but is used solely to hide the default avatar mesh only.

What this second new Alpha Layer would allow is a way to hide the default avatar completely or even just partially. So say you like a particular custom mesh body but hate the head and you prefer your LL default avatar head.  On the original Alpha Layer you could hide the custom avatar head and on the new Alpha Layer you could hide all the body except the LL default avatar head.

Another great advantage to such a system is for mesh clothes.  If you make a pair of say mesh pants with the UV layout of the LL default system avatar mesh your pants now can be textured with all the different pant textures that the system avatar can wear.  Same goes for tops, skirts and dresses.  It could really revitalize the texture industry in SL.

The current state of the mesh clothing for texture makers is they either have to pair up with a mesh maker or learn to make mesh themselves if they want to be on the cutting edge of fashion.  With this system they can continue to make great textures using the LL default avatar UV pattern and breath new life into their business.

It also means mesh creators don't have to make textures for their mesh, a task many mesh creators don't particularly enjoy doing, or are good at.  So by using the LL default avatar UV pattern and allowing the customer to wear other textures made by other people the mesh clothes becomes way more versatile.

If I was a texture artist I would watch what mesh clothes is selling the best on the Marketplace that is using the new system.  Then quickly make an awesome texture for it and put it up on the Marketplace.  This would really boost the sales of the original mesh because now there texture options for it.  A so so piece of mesh can be made to look awesome with a great texture on it.

In my mind it is a WIN, WIN, WIN !!!

Mesh designers sell more.  Texture designers sell more.  The customer gets more variety and versatility out of their purchases and it is a much easier system to use.  Just right mouse click on the mesh item or system clothes, tattoo or skin in your inventory and simply wear it.

Oh I almost for got.  It will cut down on lag due to not needing onion skinned avatars or massive scripts for appliers and huds.

[Cathy Foil]

Hey Medhue,
I just has an idea that might make it easier or quicker for you to convert your original animations using the old Bento skeleton to the new one.

What if you open your animation in Blender and then import the new Bento skeleton and constrain it to the original Bento skeleton?  Once constrained you should then be able to export out the animation of the new skeleton.

Hopefully you only need to constrain the new skeleton to the old one once to be able to export out all your animations you made.

Hope that helps.
Cathy

[Medhue Simoni]

---

Aki Shichiroji wrote:

I should note the COG isn't any higher or lower than normal - it's just that some bones are much higher than they would normally be with an average SL human. 

 

---

Oh, Ok, I get it.

If it works, then it works, and outside of unsupported ways, there is no arguing against something that works. That said, when I think about doing a ridable, I'd be very weary about not keeping the COG with the human avatar, but maybe I'm being OCD about it.

On a side note, I'm wondering where your voice indicator is on the ridable set up? If the voice indicator is directly influenced by the COG, and the COG can technically be placed anywhere, that could be good to know.