How to model such a real looking Half Life 2 main hero?

[chuawenching]

Hi there,

I am quite suprise with how the half life 2 main hero face and the entire model.

It looks so nice and convincing.

Any tips of doing that?

What kind of software did valve used to model that?

And how many polygons is that model?

Thanks.

Regards,
Chua Wen Ching

 

[spin]

Valve uses XSI for their models, but any decent 3D package can produce similar results. It's all up to the artist. The only tip I can provide without looking at a model and helping with it is to keep trying. There really is no magic bullet, you just have to keep working and eventually it'll just click. You'll learn how to create proper form and flow with time and practice. As for the polycount, If my memory serves right, the main characters are around 7,000 triangles. I'm sure someone else will have a definite answer, though.

 

[Aike]

Use Normal Maps!

(and if you need to know wot they iz... its like baking high detail onto a low detail model ... gotta lurve it)

 

[mindless_moder]

valve uses xsi to make a lowpoly model and then they smooth that low poly model and detail it to normal mapp the lowpoly model.

 

[EVIL]

wrong. they paint the normal maps. not model it. wich doesnt make some things very convincing but sinse it is hardly ever used it doesnt really mather.

 

[The Dark Elf]

It's more likely they model a high poly version and then bake the detail, bump, normal, color, diffuse etc. onto a UV map on a low poly version. We do this and many companies in the games industry have been doing it for a while. The larger studios skip the directly high poly modeling stage and simply scan in a 3D copy of a clay model, clean it up, paint its surfaces, then bake those surfaces to a low poly version, obviously I can't quite see home mod makers going to that kind of trouble (which can also be done in the same way for world textures), but there is huge amounts of detail are possible. At a price

 

[nekkidman]

jeeze, dont listen to mindless_moder or EVIL. they are wrong.

For both Doom 3 and and HL2 the process is as follows:

1) create your model (wheather weapon, character, or prop) in as high of detail as possible. you can use poly's or subdivisions or whatever. 100k, 1 million, it doesnt matter how high the polycount is, just make it as good as possible.

2) create a low poly version. how many poly's you make this model is going to depent on what kind of mod situation you have. maybe 2 - 7 k about for a character. maybe 1 - 3 k for weapon 1st person and hands. and 500 or so for 3rd person weapons. of course you could go higher or lower than any of these number if you want.

3) create normal maps. yes you could paint them but that would be tedious and probably look like crap. This is done using your app of choice. xsi has it built in. maya has some plugin kind of shaders, 3ds has something too im sure. and there are also standalone ones. i've been using one from ati's developer site. Once your low poly version is UV mapped, the high poly's surface info is "baked" into the low poly one. The creation of this map is handled by your app.

*most models if they are simply "smoothed" will look crappy in a lot of areas
*never heard of people painting normal maps, would be difficult to achieve nicely

 

[The Dark Elf]

Just to add, anyone painting normal maps in photoshop and using nVidia's PS plugin to create normal maps will find their results don't come close to baking normal maps from actual 3D data. It's just a limitation to doing it that way as you don't have any Z depth to a 2D image, whereas in a 3D scene it contains the Z depth which helps improve the final normal map.. though the photoshop method is great for veins and wrinkles where there isn't a need for a great depth to the surface - as modeling things like that (if not using a scanned clay model) would be too much work and too time consuming.

 

[The Dark Elf]

Originally posted by Fenric1138
Just to add, anyone painting normal maps in photoshop and using nVidia's PS plugin to create normal maps will find their results don't come close to baking normal maps from actual 3D data. It's just a limitation to doing it that way as you don't have any Z depth to a 2D image, whereas in a 3D scene it contains the Z depth which helps improve the final normal map.. though the photoshop method is great for veins and wrinkles where there isn't a need for a great depth to the surface - as modeling things like that (if not using a scanned clay model) would be too much work and too time consuming.

Decided to do a quick example. A note about the pic, the manually created normal map was done by creating an image the way you'd do a bump map, black low white high. But as I mentioned a few posts ago, this doesn't contain enough information and so the results are poor for anything more than small details. You wont be able to fake semi realistically deep doorframes and doorknobs and so on (from a direct angle) by using the drawn normal map method, but you will be able to fake it to a degree with the baking method. Obviously normal maps are still only 2D images at the end of the day, so sharp viewing angles will often ruin the illusion. But for the most part you'll probably be fine. Useful things for normal maps. As mentioned, door frames, can help add that little extra realism. skirting boards, dents, veins, irregular surfaces, rough ground etc. wrinkles in skin, folds in clothing, clouds (yes clouds) ripples in water, fake keyboard keys, use a normal map instead of modeling every key for better appearance and no speed hit. Faking other small changes in geometry so you don't have to model it all in Hammer. Oh and of course its going to be VERY useful for HL2's 3D skybox feature (those great looking backgrounds) You can use plenty of normal maps there and not worry about losing the illusion cause you can't get to them to spoil it.

anyway hope this is of interest/use to someone

Normal maps rock

 

[Razorwolf]

Could someone link me to a more in-depth explanation of this 'normal map' and 'baking' thing?