|
|
Post by khash on Feb 6, 2015 9:15:37 GMT
Hi everyone! I have (as in topic) good and bad news. Bad news are that development of Honey 2 would take much more time than I expected before (I thought it would take just 1-2 weeks of development after I invent a way to resolve standing issues. The fact was that research and inventing could take any amount of time but sole development should be fairly fast) Good news are that I have finally invented last bits of the Honey 2, and its crazy! And when I say crazy then I really meant it! If nothing stupid shows up to stand on my way new honey will not only be great for resources as planned before, but as well will traple previous quality with new AAA standard. It would as well make it easier and more intuitive to develop assets than before. Sad thing is that you have to wait not weeks but months for this to come. Anyway I'm so excited and couldn't resist to share  ! |
|
|
|
Post by wyliam on Feb 7, 2015 2:15:45 GMT
That's awesome news! Appreciate the honest heads-up.
Are your feature targets for Honey 2 somewhere public?
Knowing what you intend to renovate or add would help our development decisions.
|
|
|
|
Post by khash on Feb 7, 2015 13:41:10 GMT
I do not think you should rely on software which is not yet released, so please do your decisions only based on current Honey status. I would do my best to make Honey 2 great and as much as I can compatible with Honey 1. Full list of advantages and features of Honey 2 would be release at the polishing stage when I will have it all working and I will be sure that there is no hidden pitholes which may break this dream  . But I have to say that if it would go as planned we may endup with better framework for Hexagonal Worlds than AAA titles (Indie Rules!)  |
|
|
|
Post by lennartos on Feb 9, 2015 8:58:11 GMT
Sounds good
What exactly do you mean with "last bits of the Honey 2"?
From the sound of it you managed to cut down complexity of the shader quite a bit to allow "crazy good graphics" care to elaborate a bit on what idea you stumbled upon?
|
|
|
|
Post by khash on Feb 9, 2015 9:13:51 GMT
I have finally found a solution which would allow to compress most of the data we need for the terrain within mesh itself AND bake blended mesh. Which means that I can use Honey 2 tech for atlased textures for chunks, but finally I can save most of what shader did in the mesh itself. What more mesh (even not tessellated) may follow terrain shape which means less artifacts and ability to display rivers(Honey 2 had no solution for rivers). New terrain would be able to finally utilize high quality normal maps and.... yeach and much more The only downgrade is that it may again require baking process which I hoped to scrap for Honey 2 in previous iteration but its for the best |
|
|
|
Post by lennartos on Feb 9, 2015 9:35:28 GMT
Aha so you decided to bake the mesh only and then only use the texture atlas to save memory / do the markers.
You wont be able to use instancing for mesh ( that would have needed displacement mapping ) but i imagine its still quite a big step up from honey 1's memory usage. ( do you have some rough test numbers here - what can one expect ? )
Do you plan on dividing the mesh into hexes ( and then grouped like rosetta style mesh or similar ) or are you still planning to be using quadratic chunks and doing the hard math of coordinates in shaders + handling hex overlap in code?
|
|
|
|
Post by khash on Feb 9, 2015 10:40:59 GMT
Chunks would no longer be quadratic as it does not have global texture which dictated this requirement /resource saving. They would be customizable in radius hex colonies (groups)
As for memory:
- Single hex by an average have about 200 vertex for good (not tesselated) shape quality.
Color - 4*4byte = 16 byte
Position - 3*4byte = 12 byte
Index - +- 6*4byte = 24 byte // an average vertex gets indexed in terrain about 6 times, but this depends on mesh
UV - 2*4byte = 8 byte
UV2 - 2*4byte = 8 byte
Normal - 3*4byte = 12byte
1 vertex would cost = 80 bytes
1 chunk of +- 30 hexes = 30 * 200 * 80 = 0.48 MB
Diffuse and Normal is being atlased so just 1 texture for global usage so we can skip this cost as its only meaningful n very small maps which does not use a lot resources anyway.
Currently we have:
1 chunk:
DXT1 for diffuse - 2048x2048 with 0.5byte per pixel = 2MB but for some reason it always takes 2 as much (4MB)
8Bit Alpha for Height - 1024x1024 = 1MB but again texture takes twice as much (2MB)
+Mesh which we had anyway +- 100*100 * 56byte (less by UV2 and Color which we did not use) = 0.56 MB
Memory usage expected new: 0.48 MB vs current: 6.56 MB
On top of it we may possibly use Mesh Compression with Unity5, but this require further research if quality would not be harmed by this. I assume precision lost should be far below visible, but you never know until tested. This may further cut mesh size to 50% (about 250kb per chunk)
|
|
|
|
Post by lennartos on Feb 9, 2015 16:49:00 GMT
Nice, thats a whopping order of magnitude improvement  Seems to me you adopted a very similar approach to what we talked about before: Hex info inside the mesh, shader much simpler as result - meshes gruped together in repeatable pattern. Since the most expensive operation was the hex calculation that would also mean much better speed for mobile. I assume texture spaced is doubled because its mirrored inside main memory for quick alt-tab support. If you only had texture once you would need to reload all textures from HDD each time app lost focus. ( default directx behavior) |
|
|
|
Post by khash on Feb 9, 2015 17:48:12 GMT
Yes, I have to say that I take a lot of input form you and as well great amount from those who contact me sometimes over emails and I try to see if it may work. But even if it doesnt work at that time I evaluate all of it again and again when new ideas come up to see if some critical problems from previous ideas could not be now resolved. And I have to say that storing data in mesh did not let me sleep few times trying to get its way to solution |
|
|
|
Post by lennartos on Feb 9, 2015 21:11:54 GMT
One of the boons of having a community for a product people beleave in Remember to keep us updated, looking forward to the possibility of larger map & rivers + roads (also changing terrain without artifacts ) |
|
|
|
Post by MikeD on Feb 26, 2015 17:41:59 GMT
Just wanted to come in and say this is a fantastic asset, easily one of the best in the asset store. Knowing development on an even better version of it is going well and seeing all these planned improvements makes me super super happy. Can't wait ! |
|
|
|
Post by MikeD on Feb 26, 2015 20:25:59 GMT
Oh two questions though, will this be a free upgrade ? Also any ETA on when it will come out ? Thanks ! |
|
|
|
Post by khash on Feb 27, 2015 15:00:40 GMT
Honey 2 would be an upgrade on asset store Honey product so no extra costs for those who own Honey1.
I do not have proper Eta for it just yet, and development time required would make me think I should aim for about Q3/Q4 2015.
|
|
|
|
Post by MikeD on Feb 27, 2015 19:35:12 GMT
Awesome ! Of course I wish we could have our hands on it earlier but it looks like it will be well worth the wait. Wish you all the best ! |
|
|
|
Post by khash on Feb 28, 2015 12:08:19 GMT
Thanks!
|
|
! 
. But I have to say that if it would go as planned we may endup with better framework for Hexagonal Worlds than AAA titles (Indie Rules!) 
