Meeting minutes
immersive-web/capture#2
alcooper: at TPAC the capture repo was brought up
… i uploaded the PR from my personal repo
immersive-web/model-element#71
alcooper: there was a disagreement with cabanier about some of the details which need resolving before it can be merged
Brandel: at TPAC we discussed what should be in a v1 for model.
… such as an image based environment map (an ibl)
yonet: i am finding it hard to hear you
Brandel: i'll go to a quieter environment
Brandel: it would be good to work out what features should be include in a v1 here is a non-exhaustive list of thigns that would be good to discuss
lgombos: thanks Brandel, since this is specifically a V1 are these attributes mandatory or optional. One where the UA has the option to apply IBLs, it needs to be optional since it an automated one is expensive but we should let developers define their own
lgombos: does this property still make sense for GLTF?
Brandel: yeah it should be fine for GLTF since it doesn't specify an environment. The images should be any format but especially ones which support deeper bit depths.
Brandel: it should also be allowed for models to include IBLs which can be replaced by the page
bajones_: IBL isn't in gltf by default there is an extension also extensions for punctual lights.
bajones_: it would be good for the system to fallback and use it if it is avaialble but it is valuable for a page to be able to replace the IBL
bajones_: what do you envision reasonable default's being?
Brandel: for VisionOS there is a system envionment based on the user's environemnt. Which would be a reasonable default.
Brandel: but it should be upto the useragent to provide a reasonable default even a synthetic IBL
bajones_: i am interested to see what you would do for a default IBL on Mac and iPhone.
Brandel: we are basing this on ARQuick look, and it's behind a flag so you can try it. It has a default IBL but it's just a plain one that is not based on the user's environment
bajones_: i agree it makes sense for UAs to do their own thing but should probably converge on something similar to avoid creating compat issues.
Brandel: if you use xcode or reality converter you have a choice of choosing between two default equirectangular images
Brandel: I agree they should probably look similar
Brandel: two other parameters we are thinking of doing are rotation and light intensity
cabanier: i wasn't sure we had a agreed on image based lighting, and why is it needed, why would you ever need to replace the OS provided IBL?
Brandel: on the apple store online we have these fancy light setup for very specific lights and simpler set ups for our non-flagship products. For marketing pages it's good to imply the lights come from the page itself.
bajones_: to be clear if the model was not embedded inside the page it would want to use the user's lighting
Brandel: yes
cabanier: I thought one benefit of this was reflecting the environment and the UI?
Brandel: yes that is one thing it can do and is one of the goal but isn't the only thing
cabanier: is there any reason you couldn't do this with WebGL?
Brandel: stereoscopic
cabanier: it would be odd if the model was removed from the page and was in your environment that it has IBL
Brandel: it's upto the user agent to choose how to handle the provided IBL
cabanier: we want to have browser where the models stick out of the page.
cabanier: and i thought that was the original proposal
ada: is it okay if one browser displays it in a portal and respects the IBL whilst it is in the portal and another browser has it sticking out the page and never needs the IBL since it doesn't fully make sense.
Brandel: this is how we envision reaching our V1 today.
Brandel: for presentation and marketting it is valuable.
immersive-web/model-element#72
ada: I think it would be cool if the model was half out of the page and normals that point into the page get the user supplied IBL and the parts coming out get the system IBL
in 2D pages people want to move the camera to look around a model, I think we should simplify the way the camera is defined to prevent people doing things which do not make sense in a stereoscopic environment
cabanier: it soudns reasonable but I don't think we should call it a camera since at least for us the camera always come from the operating system perhaps something like "orbit controls"
Brandel: I don't like camera either and would love ot bikeshed what it should be, perhaps view of perspective or vantage?
bajones_: I think having some mechanism for directly setting the matrix if you know what to do, but also let you do simple controls like zoom pivot etc.
bajones_: it would be good to think about the default user interaction model, such as an inside out view where this a skybox
Brandel: pan and dolly would be handy as well as pivot, the issue here is to specifically add a pivot point so that you know where the pivot is. I would prefer to use the interaction model upto the UA due to differing inputs hands vs keyboards vs mouse etc
bajones_: I agree that the exact methods are a UA choice there probably will be de-facto standards will develop but new interaction models come out with relative frequency but we should plan for some interaciton modes. E.g. should i be allowed to pan wherever or rotate wherever?
bajones_: brand safety probably want to ensure people cannot see where they are not supposed to look.
bajones_: we need to balance the JS interactions with ones which can come from the OS since if we constrain the OS ones we limit the amount that can be done with JS
Brandel: I am sympathetic but I would like to leave the door open for JavaScript to handle that control.
ada: I don't like the model based on cameras since it doesn't actually make sense in XR, having a movable model on a rotating movable stage makes more sense to me. So essentially you do the opposite to the camera transform for the same effect?
Brandel: yes, I think that can be used to do the same things that camera controls can do.
<yonet> Thanks atsushi
thank you!
I was looking up the bot commands lol
i can never remember even after all these yeards