I’m trying to add 3d text to a wall in a video clip (camera doesn’t move). But it doesn’t look real. I added some blur and have two lamps (sun and lamp) in the scene. But somehow, it should have noise and something else to look more real. Some tips would be helpful…
PS. There is also another worry. The lightning of the video camera footage changes a bit during the scene. How to make the blender lights go darker/lighter during the scene?
The material you use is very important to the final look. You might try mapping a real photograph of some texture to the object instead of using Blender’s built-in materials. You can control the lighting the same way you control other animation parameters; by setting keyframes and using IPO curves.
You might want to change the lighting a bit, mess with specular, or maybe eliminate it all together. And I think a texture like clouds, but very noisy, would look good. Don’t put it on too strong though. To change the lighting just hover over the material panel for lights (F5 while a light is selected) and hit the I Key to set one. It is really easy really. Hope you get it all looking right!
I changed the other lamp to hemi and added only-shadow spotlight (three lights now). I also added a plate behind the text and noisy cloud textures to both the objects…
Attachments
UV map it!
hmm, it looks weird because you have what looks like a metal, futuristic sign composited over an old wooden building. make the sign fit the picture better. also, yes your 3d stuff needs some grain to match the [DV?] plate.
Add some dirt to the building as well, something like that sign should have some dirt running underneath it in streaks. Don’t forget AO or even AAO.
[EDIT] actually, go to www.videocopilot.net for some first-class compositing tips, you won’t regret it.[/EDIT]
Dan
Actually, it is a sci-fi film. They’re looking for parts of their wrecked space ship in a small village… so there must be something futuristic, although they’re in some older civilization. I try to add some dirt and see how does it look…
You are not compensating for lens distortion. If you know what lens the footage was shot with, match your blender camera lens to the same value. Lenses tend to “bow” out near the edges of a photo, and your sign is not.
Also the aspect ratio looks strange to me. You may be rendering square pixels from Blender against 0.9 footage.
I know, aspect ratio is at the moment the smallest problem, I’ll fix it later… But yes, I’ll change camera properties. That was a really useful tip. And a question: How to add grain, to match the DV plate, like dan_hin said?
You could simply render it out as a lossy jpeg sequence to introduce some noise. I’m not sure if Blender has an add noise feature in the sequence editor. A photoshop batch could process a series of frames as well if you don’t have a compositing program. If you do have a compositing program simply add grain in that application.
also use a cloud texture on your lamp
My main tip. Things aren’t prefect in the real world, so make sure you add imperfections.
Also, try to blend it into the surroundings. For example, if one part of the wall is more rusty then the other, extend that to your 3d model. However, the biggest problem it looks like your having is matching it to the background, which just requires a lot of tweaking if you don’t have tracking information (camera lens, dimensions of wall, etc…)
A new version… I made a new sign for the text and changed the camera lens to 37mm, like in the camera which we used in filming, but it didn’t help with distortion: Distortion still doesn’t exist.
And another problem: I tried to import the background video to blender as DV, but when I did so, the video was all black. So, doesn’t blender accept DV? So I had to convert the video to uncompressed avi and that’s why I have now problems with the aspect ratio… 
And still another problem: When I use ambient occlusion, the shadow below the sign disappears, as you can see! Why?
Attachments
- the image looks convincing.
- use the VSE to convert the DV footage to something like an avi-jpeg or frame sequence, before using a background. When you convert, use the same aspect ratio of your output as you did when filiming, usually 4:3
Situation in images 1 and 2 works (lens distortion node in “1” and combining video and 3d in “2”), but image 3 doesn’t (both lens dist. and video) so what’s wrong?
Attachments
The node has a bug. It’s outputtting the RGB data but the alpha is uniformly being set to zero. Easy fix: duplicate the node and plug the render layer’s alpha channel into the second one. Plug this into the alpha socket of a set alpha node and the result of the first lens node into the set alpha’s image socket. No more tears.
Whoa, thank you very much! Now it works! Here’s the newest version of the scene, with and without ambient occlusion. I also added some blur… Take a look. I think that there is still something funny with these pictures, say your opinion. And I still can’t understand why the shadow disappears when I use ambient occlusion…
Attachments
Include the shadow pass in the render layer then use a mix node to multiply the entire scene by the shadow pass B4 you output this mix node into the composite node.
If you look carefully the shadow is not entirely gone, it’s just being lighted by AO to the point that it’s barely visible. Remember: AO is basically a “Sky Dome” so you’re getting light from 180 spherical degrees.
I think it still looks a little fake to me (although the AO render looks best, that shadow does not match the lighting of the scene). The reason why is that you have spent time distressing the sign, but not the barn that the sign hangs on. Maybe make one of the letters in the sign crooked and add some rust stains on the barn wood from the nails that hold up the sign. What is holdingup the sign? How is it attached to the barn?
Another thing - you might want to add a little rust or some leakage/mold/rust from the sign onto the building.
huge improvement from the beginning
To add noise to it you can do that in the compositor and get it looking pretty good with some experimenting.
>You add two or three texture inputs with different noise textures on them - change their colour in the colourband settings.
> Add a time input and use that to change the texture.
> Then ‘Mix’ them together - add a soft blur possibly and also use RGB to change intensity.
> ‘Mix’ again with ‘Add’ (noise over video) and change the factor about + RGB curves.
Engish is my first language - sorry if that makes no sense.
if you want a blend example let me know.
Hope that helps -
miff