Anata VRM Call 12-5-22

https://hyperfy.io/anata
https://github.com/M3-org/anata/issues/12

Checking out Anata Figma
https://www.figma.com/file/IQMTlh0ofxhUju5I7pQMPA/Anata-VRMs?node-id=0%3A1

traits scrape

• can we do a (?) of shared traits and traits that are specific to gender

  • list of unique items
  • compare shared traits with masc and fem; create list of uniques

• guilty: issue with some traits being called the same thing but represented with different versions

  • on situations like this make one main version; revisit variations later on

• trying to get trello access and ppl added;

this week

• hongzi: wings and head decoration, only female specific

  • jin: we should cross check which ones are already done
  • jin: guilty do you have them tagged to know which traits have been modeled
    • guilty: not specifically yet. when we started artists we only had depository of images; they've just been going thru those, not sure which traits they align to in anata files. our artists are just calling it hat1 hat2
    • jin: it would be very difficult if we had to cross check everything after they get modeled to trait names

• guilty will be matching what his artists are doing to what the trait names are; more of a manual task

  • jin wants this to be done going forward; i.e. keeping name and models exact

• jin: GLBs for modeled traits

• hongzi: what is best way to share

  • jin: blend file

• guilty: plan for final exports?

  • hongzi: assembly pipeline in figma?
    • jin: still working out assembly but in process. for now work out process to prevent duplicate work and keep well sync'd. keep things – good project management

• hongzi: will you guys make those traits – do you make sure the size fits exactly on the model? like on the hair – they fit exactly. it won't be like the head penetrates.

  • guilty: i dont have any of the stuff you guys have worked out.
  • hongzi: do you guys do sizing in assembly step?
    • jin: differences in size and rotation per base mesh. there will be some general resizing of things that will be necessary. once models are generated will need a variety of tools to position them in a master blend file
    • gulity: create bases; distribute copy to artists; artists model and fit everything to that base model; export just the trait or part that they worked for

jin: generate unique male / female lists from diffs to shared assets
jin: guilty - tag each completed wearable with trait filename, share asap

• jin: masc base?

  • guilty: not done yet; wasn't sure
    • guilty: we will make male base
    • guilty: would like copy of female base to make sure it fits
      • jin: good idea to have this

• hongzi: i was looking at github and sub tasks –

  • jin: wants to get it sync'd to dework

• hongzi: how do you guys check the work you are working on this week. is that on dework or somewhere else

  • jin: the best way, what i've been doing is putting them into figma, update hackmd with figma stuff. figma's our huge white board to always see big picture.

• hongzi: we can do whatever you are most comfortabe with. what we are doing – its not exactly week by week. its just whenever (inaudible) … as progress moves forward (inaud) … we can have artists come up with better lists

• guilty: (inaduible) something about dework

  • jin: waiting for a feature where everything gets imported as subtasks; for the time being i say we keep it high level as to – okay if we are going to work on this then lets make sure that the other team doesnt do the same work so that it isn't overlapping

• gilty: lets get on the same trello, that would be helpful

  • jin: i just joined; i could import trello into dework and we can – it will maybe take a week or so. dework will sync a source of truth for both. for now just being invited to see the artists trello, that would give us visibility to check and reference to make sure we aren't doing same assets at the same time

• hongzi: a lot of our artists are asian timezone. when they wake up we can get ppl added to trello

  • jin: good to get us all added there but we wont migrate all of this all to trello, use it for ref
  • jin: let me see how dework trello integration is – we will know by tmrw.

• hongzi: loading this model into hyperfy it says very poor performance. is there any refernce of poly count we want to keep in mind?

  • guilty: in my head these models should be able to use anywhere; the tighest budget i can think of is vrchat on quest which has i think
  • jin: we do have optimization pipeline. we have tools to streamline creating lods and everything. its better to start high and go low rather than go low and enhance.
  • jin: per wearable; i dont think jackets or headgears need more than 4k. or – its fine if you're 3k. or 5k. i'm good at optimizing, we have tools to optimize

• hongzi: we wont worry about poly count right now; optimize later.

  • jin: yes this is fine. material count huge. pref 1 material per object
  • guilty yes this is huge

• guilty: with performance its not usually poly count, its more about draw calls.

• hongzi: do you have system to optimize texture file?

  • jin: just general optimization but also ktx2 basis compression. can get very very good savings out of it even – workflow is not very artist friendly right now but a lot of platforms support it.

• guilty: avoid level 5 subdivided stuff

  • lol (everyone)
  • guilty: a gripe of mine. NFT collections boasting "game ready" assets and its millions of polys + 30 mats; crazy – something i really want is just super optimized. plug and play, performant, can use it anywhere

https://docs.vrchat.com/docs/avatar-performance-ranking-system
^vrchat avatar optimization standards

---

Action Items This Week

• guilty: tag wearables already done with traits
  • can send files to Arashi / jin to help with this
  • From now on keep project filenames for each trait consistent with metadata
• hongzi: diff shared traits to male / female to generate unique lists
• hongzi: transfer female base mesh to guilty
• guilty: design male base mesh in sync with female base mesh
• hongzi: invite me and guilty to trello