SDK2 or SDK3 for beginner at world building?

I’m a total beginner but I spent some time trying to learn before asking this question.

I’d like to try building some worlds - nice, but maybe not groundbreaking. My background: some very limited visual basic and I’ve learned to build a few very simple SDK2 & SDK3 worlds (a disc, some imported models, shaders, skybox, few lights).

My question is, should I start with SDK2 or jump into SDK3/Udon from the start, even if I don’t think I need Udon’s capabilities at this time?

SDK2 - Definitely Bakery and Mesh Baker are compatible. Zillions of Youtube videos. It just looks easier. But maybe I’m wasting my time learning something that will become obsolete, or that I can’t update with some new tool to better handle VRChat visitors?

SDK3 - Future proofed but I don’t know about compatibility with Bakery/Mesh Baker and I’ve been told that some asset store landscape & other kits won’t work because they’re script-based. Far fewer help videos and pre-fabs. Most important - learning Udon and far worse, possibly C#…looks like a big and scary time commitment, especially compared to SDK2.

Any advice is appreciated!

generally I’d say in terms of avaliable tutorials and information on how to do stuff SDK2 is well documented, where as a lot of SDK3 world builders tend to be Udon Sharp users. There aren’t many graph users around. While there is some documentation around for those who are using both methods. Generally more advanced graph features aren’t covered much. So generally the best thing to consider is while SDK3 has much more options and potential in the world you can create, you may need to consider delving into C# at some point.

Though I’ve managed to avoid it so far as someone who isn’t as C# savvy. Just a lot of graphs take a bit of out of the box thinking. There’s also some great emulators out there to test your works quickly as well. Though as someone who’s been using sdk3 for the last 6 or so months now. I find it difficult to justify why I’d want to go back to sdk2. I believe it’s due to the ease in re-usable behaviours and just the overall more in depth control I have. I still think it needs a bit of time to work out some minor issues but it’s improving since release.

Thank you! Thank you very much. I really appreciate that, you’re exactly the kind of person who could have best answered that question. Udon sharp does scare the bejeepers out of me so it’s good to hear that you’ve been able to make do with graphs, by applying some extra tactics. I admit to being pretty enticed by the idea of greater control and being able to incorporate work that people will do in the future. That’s interesting about emulators, I’ll look into that. Since I wrote the question I’ve sort of come to the conclusion that a lot of sdk2 material (like models for example) are going to come over fine, and it seems like some key tools like Bakery and Mesh Baker will be ok also. I think your post is going to help me take the jump.

GitHub - CyanLaser/CyanEmu: CyanEmu is a VRChat client emulator in Unity. Includes a player controller with interact system. Works with SDK2 and SDK3. Give this emulator a look at, a few of my co world developers use it and honestly props to Cyan for making it.

THanks (for doing my homework!) I’ll absolutely look at that.
Maybe I can return the favour, I also came across these on Orel’s discord server, they are beyond my ability right now but maybe something for the future.

Console Enhanced Pro (there is a free version also) on the asset store:
https://assetstore.unity.com/packages/tools/utilities/console-enhanced-pro-11521

Enhanced Hierarchy 2.0
https://assetstore.unity.com/packages/tools/utilities/enhanced-hierarchy-2-0-44322

His comment was “can’t recommend enough”

Hey, thanks for posting this. I’m in the exact same boat and just assumed SDK3 with graphs was the real option (I use the Avatar SDK3). Being a visual artist, I think I’ll look at not punishishing myself with C# and try SDK2 instead. I have a few casual games up my sleeve for people who want to seriously play games and not chat all day.

I’ve built world’s in Altspace, but as far as I can tell, they don’t have a scripting layer. So it’s all beautiful but non-interactive experiences.

I’m confused my VR Chat devs don’t just hire pro developers to build sticky worlds. Everything is left up to JAR and donations to build.


From my perspective, I’ll be taking a long, long time to work out how to build, but maybe just important, to understand what I want to build. I’m interested in a creative outlet unrelated to work - so I’ve got time. It might be genuinely more satisfying to slowly learn graphs or (gulp) some limited Udon sharp/C# and really understand what it is that I’m doing. As well as providing the potential advantage of future-proofing.

Why doesn’t VRChat hire pro devs to build “sticky” worlds? Sounds like a great idea, along with a thousand other “why don’t they” questions, from technical to marketing to customer relations. Always easier to be an armchair coach I guess. Maybe they’re focusing on building blocks first - like developing Udon and (perhaps) slowly assembling a more stable and hardened platform. For example, I’d suspect their resources have not been too heavily invested in a decent marketing plan! Some of the opportunities that have been missed… And my profession isn’t even marketing.

As a casual user, I just happened to stumble on the Jean-Michel Jarre New Years concert. It was brilliant, one of the best events I’ve attended , news about that should have been slapping me in the face for weeks. I’m assuming that, just like in my own work, there are a thousand opportunities to make things better but just not the resources to get to them yet. After all, this isn’t Microsoft - I’ve heard they only have 15-20 employees. Hard to build a new industry and new paradigms, as well as get the office toilets cleaned and the floors washed, with that incredibly limited amount of resource.

Hopefully it sticks around long enough so that it can begin to realize it’s potential. There seem to be so many opportunities.

SDK2 is very well documented, but a lot of interesting things in sdk2 require EXTREMELY round about methods & deprecated unity components & old assets to achieve, where in SDK3 with a little bit of digging on youtube you can find simple Graphs to get the same things done. Just as a heads up! “well documented” may not always necessarily mean easy to figure out especially if something is 2+ years old & there’s nobody around to ask about it.

As someone who messed around with map making in sdk2 & sdk3, knows C#, has used U# BUT exclusively uses Graphs for udon projects, large and small. I think, if you plan on doing anything that will be interactive in a map, then SDK3 & the Graph system is worth learning the basics of!

It may not be well documented but most things done in U# can be done with the Graph system. I’m not sure if most people just have an aversion to node based systems but Udon Graphs are pretty good once you learn enough to get creative with them!

@WoogyT Thanks for the thoughtful and detailed comment. That’s very reassuring. And sounds like wise advice - you hit one of my concerns, that if I’m going to do this slowly over the long term, taking the SDK2 route could lead to some future disappointments. Looks like you’ve posted some interesting graphs yourself and I’ll follow them once I understand more. This is definitely the route I’m taking, Udon Graphs, as a result of the first post from nerzarn and now yours, and see where I can get with that.

Also, given that this post has about 100 views now, maybe there are more than a couple of us who have been wondering about this question… Thanks again!

1 Like