Hi all.
I have the last 6 moth spent a lot of time learning blender. It is still a long way to go, but I am finaly starting to get a hang of it
The last month have I showed some of my work to my co-workers and they are starting to realise that they have a 3D artist (well still learning 3D artist) under the roof.
So i start to get a chance to do a work we usually would by of someone. So here is my secon job. It is a logo in large sice where we later on will photoshop a leaning person to it.
I still dont think it looks real enough. So I thougth you guys maby could help me in giving me tips on what to do. I know the raytraced shaddow might be a little to blurry. I will work on that. It is not ahuge picture. So if it is to small give me a shout and I will render a new one ( it took hours to reder this one so please dont say you want alarger one )
first of all i think it looks like the logo is hovering.
maybe u should try adding the lady to a plane as an alphatexture(i´m not sure if it´s gonna make it alot better though:D)
try rendering with indigo too, it gives much better results
It looks pretty good so far, fairly simple (Comparitivly of course) the colors do seem to be a little faded, IDK if you planned it that way or if it’s your render settings or what. “Hours” seems like a little much for a simple render like this, I would check with one of the more experienced guys (or gals) to see if one of your render settings is out of place or something to speed up render times.
First of I like your version better then the sketch but that is besides the point.
The angle of yours is different from the sketch try to align it more.
Your lightsource seems to be stronger and your material has more spec then the sketch.
The lightsource in your render also seems to be coming from a different angle then the sketch one. See the inside edge of the L standing on the groundplane. In yours it is highlighted in the Sketch it doesn’t receive a whole lot of light!
I would refrain from getting the shadow to look like your sketch but go the other way. Try to model a cutout from the model that casts just a shadow on the ground plane. Later simply plugin in the model with photoshop over your cutout and voila the shadows match up! You might also try this in the blender compositor but I’m guessing that doesn’t fit into your pipeline.
First off, the sketch is just a sketch. It doesnt need to look exactly like it (sorry for not saying that). So the different in where the light source is, is totaly ok. As long as it looks good.
Thanks for the tip with shaddowing only acut out. Thans some good stuff :).
When it comes to rendering time. Yea, i must have something set wrong. is it ok to mix buffered and raytraced shaddows in one render? Becaus I have dome that, and thats the
only thing I can come up that could mess my render time. But I am all ears!!!
musk already mentioned that the apparent camera angle of your render doesn’t match the sketch, but one other thing to keep in mind is that the perspective of your logo model and that of the person will also have to match closely or the composite will look false. To do that you’ll need to match the camera field-of-view and camera position of the person-shot, perhaps not exactly, but it should be close. Is the person-shot something your agency will take, or a stock pic?
To match the real-world camera lens specs, the BLenses script can help.
Keep the settings for the actually used shot around in order to use them with the Blenses script or with another renderer that supports camera lenses settings.
Why are you using raytraced shadows at all? Since the image isn’t overly complex I would not use raytracing but instead go with spots and buffered shadows all the way. This way your render time should drop to a minimum. Use AO for lightning your model additionally.
Since this is a production shoot you will probably not have this option but you might want to install one of the latest SVN builds from graphicall.org and use the new Approximate AO since it is way faster.
This build also gives you soft shadows for all raytraced lamps not just arealamps but as mentioned before I’d try to go without raytracing.
Record the position of the camera in relation to the subject as well – height above ground plane and distance from camera to subject as a minimum. Then in your .blend, set up the scene as if it’s a scale model of the live-action shoot. If you then use BLenses settings that match the camera/film combination, you should get a Blender camera field-of-view that will help the real-world and virtual perspectives match closely.
Do what chipmasque says, and everything’s gonna be fine.
If you don’t want to install and use BLenses (i strongly recomment that you learn that, if you are thinking in compositing with real shots), maybe you can check with the photographer at least the Lens that he’ll use, so you can set the same in the Blender camera.
Not quite the best route, because the native Blender “Lens” setting doesn’t correspond to a focal length in mm. That’s one reason I wrote BLenses, and that’s its main purpose. It has bells & whistles as well, but the basic functions are pretty easy to use.
If your photographer knows how to calculate a field-of-view from his shot setup, you could use that in the “Angle” setting for Blender’s camera. BLenses automates that calculation, though, and may be simpler in the long run.
The angle of the human model in the image is totally wrong. You have to get her back in and redo the angle and her pose, she looks like she’s nudging the logo, not leaning on it. Is that stock photo or from a shoot comissioned by your company?
The composition of the shot should be decided with the human model first and to have a reference mock logo (a section of wall, not a complete reconstruction) so that the angle of the rendered logo is correct, that has to be right from the get go. Second is the pose of the model, to be more relaxed in the shoot so that she is more casual but still confident like she is in the sketch, her arms still crossed with her feet crossed over at the bottom like a 4, her inner foot pointing to the ground while her outer foot supports her body weight against the logo. When you have that all sorted the logo has to be correct in lighting so that it matches the photo. Use a reference like a white ball or cube during the photo shoot to get the right fall of light on a solid object. Most likely it will be diffuse light but get the right ratio of shadow on your logo. Render it in Indigo if you can. And have a floor rendered underneath it too with a white background so it’s seamless with the floor. The shadow of the logo should take care of itself. Fill in shadow detail in photoshop for under the model.
Wow… lots of great replies. But it seems like some of you think the picture with the girl is anything I have done. Thats not right. It is a really fast sketch by one of my co-workes. Just so that I could understand what they are looking for.
Ok back to topic.
To begin with, thanks chipmasque for the BLenses. It will help a lot.
Musk: I actually started with bufferd shaddows but I didnt get the shaddows i wanted with it so thats why I started to fiddle with raytraced shaddows. But It sounds like I maby should give the buffered shaddows another try. Especially now that I got the tip of rendering a cut out. I think that might produce the shaddows I want.
With “Ao” do you mean Ambient Occlusion
To render the logo so that it looks like it is standing up, you may have to use a longer lens (like a 50mm, not the default 35) and you want to ensure that your vertical lines are true vertical (up and down). So check out architectural rendering in the wiki to see how to use the XY offset in the camera settings.
Find out the height of the camera when they shot the model, measure the height of the camera so you can put that into Blender. It is usually eye level. You also need to note the location and type of lights they used, so you can match that in Blender. They eye is very sensitive to lighting changes and camera angle differences when merging CG and live-action.
No, those don’t help. What is going to make a big difference in render times is the lighting set up and the materials set up.
Which lights are you using, what are their shadow and spot settings?
What are the shader settings, the mirror/transparency settings on the materials you’re using?
Also, do you have ambient occlusion turned on? If so, it’s settings also affect render times.
Orinoco: The materials are realy basic with some bump textures.
Two lights are using buffered shadows and a third uses raytracing with the third is a square area lamp with raytracing set to 8 samples. ( it was after that I changed to raytracing that it got slow) As I said I will try to get the shadows I want with only buffered shadows.
By the way, when should I use buffered shadows and when to use raytraced shadows???
I thought the raytraced shadows are always better looking than buffered!!?? But if it takes this long to render ray shadows, then I will not use tem often
Not quite the best route, because the native Blender “Lens” setting doesn’t correspond to a focal length in mm. That’s one reason I wrote BLenses, and that’s its main purpose. It has bells & whistles as well, but the basic functions are pretty easy to use.
Thanks for the correction. I know that difference, but for this project, if there’s lack of time it should work, dont you?
To render the logo so that it looks like it is standing up, you may have to use a longer lens (like a 50mm, not the default 35) and you want to ensure that your vertical lines are true vertical (up and down).
i think that convergent vertical lines make the composition looks better, like the logo viewed from below; that makes the logo more “important”. Just my point of view.
Two lights are using buffered shadows and a third uses raytracing with the third is a square area lamp with raytracing set to 8 samples.
I think that the better option is to use one or another. I believe that using buffered shadows will make it faster to render, and using AO too will achieve some good results on realistic shadows.
That would do it. An area lamp set to 8 samples is the equivalent of 64 point lamps. Each ray traced shadow adds to processing time, and you added 64 of them.
Buffered shadows are probably what you want to use. Ray traced shadows are very accurate, but have hard edges. Take a look at the wiki article on buffered shadows to get a feel for the different settings you can use.
)
