The 1 Hour Gallery
A Nokia WAP
Okay, the pic talks for itself... really simple model, but Highrez textures.
I scanned the phone at 150 dpi, and used the images both as a reference and as a surface (I didn't even modify them for bump, but used as is).
I would usually have modeled it better, but the schedule was REALLY tight. It took about an hour from scratch to F10, including scanning.
I started with a subdiv box, metanurbed it, and started pulling points.
BTW, i just love L6's modeler - i had the textures applied while modeling, so i could clearly see what's needed and what's not.
Here's some notes on speeding up the workflow in emergency situations, from one of my posts to the LWML:
Well.. i'm sometimes willing to sacrifice the quality a little in order
to meet the deadlines ;-)
Yep, faking is the name of the game.
Only model what is really *seen* on the screen - why model a beutiful
backyard that's obscured by the building in the shot?
Decide what's important in the scene, and focus your efforts there.
No-one will notice the accurately modeled threads in a screw that holds
the light knobs in the upper left corner of a room scene, if there's a
cool terminator model in the foreground.
Actually most of them won't even notice the screws.
Mmm.. come to think of it, some will not even notice the whole light
knob... or the room...
In many cases the backgrounds can be image maps or rough geometry with
good surfacing. Instead of modeling the lightknob in the above case,
take a picture of one, and add that as a decal in the wall surface.
If you map an image to a polygon you can move and pan the camera and
still make the movements match with the foreground. This can be ie.
photoshop-enhanced render from your scene. (for example render a few
tree and copy them 1000 times in PS)
Or, better yet, make an environment cube or ball of your scene.
Even relatively close objects can be left very rough in some cases - see
an example here (Nokia WAP):
Theres not much detail in the model (subdiv cage), it's all in the
Recycling is one factor that greatly speeds things up - it's often
faster and easier to modyfy an existing model than it is to do one from
When it gets hectic, I often cut down render times by rendering
background at 1/2 resolution and blurring it a little - I get DOF and
speed at the same time ;-)
Same works fine for hypervoxels and volumetric lights.
Also, you can separate the elements that need a lot of AA, and render
them with higher settings than the rest of the scene.
It doesn't always have to be 3D - many effects are faster and look
better in 2D.
90% of the shots i make, and end up being used, are not pure lightwave
renders, there's usually some color correcting, blurring, graining,
blooming etc. happening - and some 2D elements comped in.
Back to Eki's home