FOnline Development > 3D Development

3d models development

<< < (459/482) > >>

Lizard:

--- Quote from: Luther Blissett on April 19, 2012, 06:36:13 pm ---That's awesome. Is that the same sort of technique (projection painting sprite on from various angles)? If this does work for base sprites of course, we have to try and work out how to get a similar look in the new added content. I do think a lot of it is choosing the base colour from a fairly restricted palette (so a brown jacket is a recognisable "Fallout brown" etc), then using a bit of interpretation to try and get a "true" version of the garbled pixel mess.

--- End quote ---

That sounds really promising! Tried this trick with a leather helmet image. Just to ask - were the helmets from VanBuren or made by community member?

Johnnybravo:
Allright I get it. Perspective and shading is irreversible so one can't perfectly reverse-engineer fallout sprites (or at least not without overkill approach).
BTW I think the more interesting is that the mess still have some shape and colors. You can indeed blend it and use it as background to do corrections.
But more interesting is that you could easily use it to make whole new texture. If you look at the tribal guy and the texture you got from projection, you could easily convert it to height map (though it'd be kind of mess without repainting it), and as diffuse use pretty much just solid color, or something including a bit more detail. Combined together, result would indeed be very impressive and possibly even much easier to create than manually painting skin.

Luther Blissett:
Lizard - I'd be really interested to see how that comes out like - you'll have to somehow compensate for the "not visible" areas, but it's be interesting to see how this affects the non-original models. If you find any good tricks, make sure to let us know. I'm not sure about the origin of the graphics for those though.

Johnny - I'd be certainly interested in any thoughts / ideas / methods you have regarding this. Even if I can get something working well for Tribal and Vault Suit, it's quite time consuming, and though I imagine the 3rd and 4th ones would be much quicker, it's still going to be quite difficult to explain exactly how and what I've done - then working out how to get a similar look for our added models and textures (I'm looking at my own combat jacket and longcoat and thinking what I can do for those).

As said, it was just some little tests for me to work some stuff out, but now you start to mention these other things, it could potentially be much more interesting and useful than that. As you said, if this "pixel shape" could be formed into a reusable height map (or perhaps an overlay template for photoshop?) perhaps we'd be able to mass-produce these effectively.

Anyway, I think on the very positive side, though still quite unrefined, there's probably enough evidence to say that we can get things to look "right". Especially when you look at the work Karpov has been doing with the animations recently. The question is how to combine this with something which works well for all other aspects of the texture and models and is easy to do. I still think metal armour is going to be a very tough one, because it will have a high reliance on the specular layer for its lighting.

LagMaster:
wow, some nice progress there mates
can't w8 to see them in game

Johnnybravo:
If you look at those projections closely, they are for sure quite crude and distorted - and this error probably cannot be processed automatically.
But you can see how clearly it mapped shaded parts of sprites that you need for recreating lightning. Converting them straight away (some kind of photoshop script or effect will be enough), should still recreate very precious details of original model.
Converting this to normal maps that can already be used fine will basically add details from sprites on those models.
The concern is how bad is that distortion and whether there is any chance to fix anything or compensate it.

For traditional vertex shading, you could use it with lighting features that photoshop boast with (and other apps as well), but it's pretty much the same thing.

You don't have to worry about specular shading much - it can probably use some adjustments in code (like changing position variables, which are just a wild guess now).
In 90s specular was almost always done by just masking intensity against white reflection.
Masking itself takes just few brushes to exclude non-reflective surfaces and color level adjustment, it can already look great after 1-2minute of work.
They did not bother with it much back in 90s either, so no problem there.

The biggest challenge is just to make low-res models competitive with high-res ones (even though they are few pixel of 256 color sprites :d).
That is the reason why I suggest using the projection for height-mapping/normal-mapping.

Navigation

[0] Message Index

[#] Next page

[*] Previous page