After some scaling and rotating of the different scene elements, this is what I rendered.

As you can see, the reflections work extremely well, you can see the signs on the building reflected on the car, as well as the parked red car.
because of the way the photo is mapped onto the sphere, the IDL lighting is very accurate, despite everything "behind" the camera being a mirror of what is infront.

In this render (and the others I will post in a mo) the intention was not to be flawless, these are "concept" renders, the car has not been completely shaded (obviously)

each of these three images were rendered one-after-the-other changing the photo(scaling the sphere, and fixing horizon pos), camera angle, and lighting angle. Changing from one photo to the next took me about 5 minutes before I was ready to render.

For 5 minutes work, these images (especially the first one) look fabulous!

Just so you understand, I am not post-processing these cars in, this is the image outputted from poser

looks fantastic.

make the AO bigger and i think you will get perfect results.

post more.

IDL will ignore AO, and I am not sure if having the AO node evaluated does any significant good.

except for the 2 tiny screw-ups in the AO under the car, ive got the shader working better, but still not great.

I called it a fake AO catcher, when really its a "Car Undershadow Shader".

when using IDL, obviously the light can get under the car, because there isnt actually a ground there (it is completly transparent, and therefore ignored) so this is my quick-and-dirty work around ;)

the reflections came out extremely well.

the next step of course is to use the custom outputs to seperate the env sphere from the rest of the render so that the car can then be taken and layered onto the original photo.  time to head to the manual :p

I'll post my Car Undershadow Shader in a mo

well this custom output to PSD layers built into PP2010 is terrible!  I dont know if you know how it works, but you plug a node into a custom output, and the result of that node is saved onto the PSD layer called custom_output_1 (or 2 or 3).

So, if you plug a diffuse node, you get a black layer with just the diffuse for that material on it.  It sounds good, and in very restricted situations the feature is fantastic. BUT, whoever thought is up must have been tripping off their head at the time.

I want to split the car from the background, so I can layer the car onto the original photo.  First I attempted to remove the env. sphere by placing it on its own layer, but as there is no AMBIENT node, you cannot have ambient levels on a seperate layer (why!?!), you also cannot remove something from the overall render onto a layer, it will still render the material onto the background layer, just ALSO onto a PSD layer.

This meant I would have to seperate the car so that it is rendered on one or more custom layers. having only 3 custom outputs is not enough for 90% of the car materials (you have diffuse1,diffuse2, specular1, specular2 at most) in order to replicate the same control over these aspects of the material, I used MATH nodes to replace the root node controls. Just a simple color_math/add node with the diffuse/specular node attached to it (two onto one of them to make 4)

the render came out with 3 completely blank layers in the PSD!  My only conclusion is that you MUST have a diffuse/specular node plugged into custom_output's on the root node for anything to show up.

I really REALLY hope someone comes along and shows me the errors of my ways, because this feature should be fantastic, but as I see it now, its almost useless. Its also so complicated to convert a scene into multi layers it would be easier two do to renders, one of them being a mask to cut the car out, and use photoshop to seperate the car from the image!