So far we have talked about a huge load test this coming summer. It will be a virtual “Woodstock” labeled as the “Summer of Hugs”… I also think that we should call it “Woodshock: Summer of Hugs. ™”
We will use container virtualization I have been working on which will deploy containers to expand the assignment clients. When the resources reach 75% on a node we will deploy another node and grow out assignment clients horizontally, we will start to grow our nodes vertically in scale every few nodes or so. I need to determine the best way to do this.
Ideally, we should be able to break the 400+ avatars in world record held by Hifi and I expect we can reach 1,000 with this type of virtualization. This stress test will help us tell where we encounter problems.
So far we have been discussing how to enable musicians the ability to virtualize their band in world. What we think will be a reasonable method is to use the web sdk developed by Vircadia and incorporate tensorflow.js to perform real-time motion capture from a phone or web enabled device. Any device with a camera that uses Chrome should work. A musician can capture their movements onto a NPC or Avatar…
Another idea is that since they are not going to use the mic to pic up audio, their sound tech can wire outputs from their mixer into the device. Then we can attach the audio to an entity or avatar to provide simulated stereo… This concept obviously needs fleshing out with real examples…
One other thing we need to make this work is some way for bands to register and take time slots. I’m thinking I need to develop a Hud App that will:
- Select a band time slot.
- Add links to NPCs
- Assign NPCs to Zones
- Attach tensorflow poses to NPCs
- Assign Audio to Zones
Using the Web SDK, the band needs to be able to logon… not sure how much of the above functionality can be incorporated but being web tech it should be somewhat flexible.
I will use this forum to discuss and record notes on the development… please share insights, ideas, suggestions. You are welcome to participate.