I don't think this has been mentioned in this forum before (although i'm pretty sure Rich_is_bored was working on something similar).. but Ryan Clark came up with a really cool way to generate normal-maps from photosource, as well as a tool for generating displacement maps and stuff.. I highly recommend this to whoever hasn't seen this method, it may give you some ideas!

Anyway you can check it out here... http://www.zarria.net

That's actually quite neat and produces pretty good results.

rich_is_bored wrote:

Ooooh. Very clever.

A logical extension of what you did with the scanner approach I bet..!!

OMYGOD that is just freakin awesome. i mean if we made actual charactors with real stuff, and used that 4 position light thing, we could make DOOM3 look like the unreal 3 engine at least with this you get some realy nice bumpmapping on the normals, great tuturial, im keeping this in my favorites

I think it's much simplier just learn to make a 3D model than buy a camera, setup objects in your room and take photos..

It's a really interesting tecnique but limited/applied only to small objects..

I wouldn't use this method for 3D objects.

You've got to deal with the fact that you can't capture the surface a whole 3D object in one shot. This means you'd need to stitch multiple pictures together to make the texture for a model.

This however, is a very nice technique for capturing normal maps for map textures.

Infact, I used it to make a carpet texture awhile back and it worked quite well...

The hardest part is stabalizing the camera so all 4 shots are consistant.

ok i see what your saying, yeah the texture wont be tilable but, yeah its good for little things like tech stuff etc, or whatever you want, but its pretty cool though, i took this rubbermaid panel that was laying around in the garage and it made a nice normapmap off the 4 light positions, just amazing that its how normalmapping realy works in there, but yeah i know how to model 3D so, but i just got amazed just taking photo's on this though with 4 shots to calculate the normals, very good effect for small stuff you wanna do in DOOM3

Actually, it's not too difficult to make a tileable texture. I did that with the texture above (look at the surface relief, not the colors).

I was actually talking about the different sides of a 3D object.

For instance, if I wanted to capture from something like this...

to apply to a model, I'd need to flip it over and take a whole new sequence of photos for each face.

Then I'd have to stitch all the resulting normal maps together so they matched the UV map of my model.

The truth is that if you can model, you might as well model a high poly object and render normal maps.

This technique comes in handy when modeling a surface would be more trouble than it's worth.

For instance, carpet? Modeling carpet would be stupid considering how detailed you'd have to be in order to get a realistic result. Not to mention it's kind of hard to visualize what carpet really looks like up close.

In that case, capturing the surface relief using this method would give you quality results quicker.

i like your thinking

can't you just do the lighting from different directions using the light effects in Photoshop?

You mean applying lighting effects to a single photo 4 times to produce the 4 images needed?

You could but you'd probably end up with a result that's similar to running that same photo through the Nvidia normal map plugin.

I'm not sure what all the fuss is. Provided you have a digital camera and a flashlight, this is really easy.

ajerara wrote:

can't you just do the lighting from different directions using the light effects in Photoshop?

That doesn't *quite* work because when you bring in the original photo ideally it has to be 'prelit' from just *one* angle; 0, 90, 180, 270. It introduces some 'anomolies' if it's not when any one of the angles is out of sync by more than 5 degrees.

not fussing, just thinking. Yes, it seems like the light would have to be just right in the one photo you would use so you wouldn't get light coming from more than one direction.

could someone please explain Step 3 to me:

"Step 3:
Create a new image (hereafter called "AboveLeft"), with your Above-lighted photograph for a green channel, and your Left-lighted photograph for a red channel."

how exactly do i make a picture an rgb-channel? i don't understand what he means at all.

kind regards
gleeful

for Gimp Users , in step3, is to take the the photo with the light from Above, and make it all Green rgb, in gimp you go into your, layers dropdown,under colors your see right at the top, color balance, in here your gonna turn off all the,, RED and BLUE , in the three different ranges, Shadows,MidTones,Highlights, and that will give you the green channel, now just do the same for the photo with the light on the Left and make that RED, in photoshop its probably alot easier, so if your using Gimp this is how i do it

thanks for the reply doomkid3000.

i think got that first part now - but he doesn't explain how he makes the 2 pictures one (the Above-lighted photograph that becomes the green channel, and the Left-lighted photograph that becomes the red channel).

what layer-merge do you use?

kind regards
gleeful

Well since his instructions are geared towards Photoshop there are some discrepancies that arise when you try to apply those steps verbatim in other applications.

What he's essentially doing when creating the "AboveLeft" image is copying the greyscale information from the above and left photos and pasting that information in to the cooresponding color channel.

In other words, because the data is being pasted directy into color channels there is no need to blend multiple layers.

I suppose the best alternative for applications that do not support direct channel manipulation would be to colorize the separate images as doomkid3000 mentions above and then use an additive blend to superimpose one colored photo onto the other.

well, i do have photoshop at work - but i must admit that i had no idea that you could paste pictures directly into rgb-channnels.

how do you do that if i may ask?

thank you and kind regards
gleeful

edit: thanks, i got it now - it doesn't seem to work if you have a transparent background.

anyway in Gimp, i use Layers for this, and i use the blend mode (Screen) and on the last step i used a (Grain Merge) mode, witch gave me the exact same results as the tuturial shows, so it can be done in Gimp aswell with the same results