I'm so excited I came across this post. I've been scouring the internet for this info all day and you gave me exactly what I needed. Are you still working on making 3D assets into 2D ones? I'd love to hear what other things you've figured out. Thanks again!
I'm surprised there aren't more responses here. I once tried to achieve the same thing. In fact it was to "reverse engineer" an old game rendered sprite to a close 3d model.
Anyway, cool investigation and nice model you have there.
nice report - i did some iso stuff in the past. For me -30 instead of -35.265 worked a bit better. And for final images "mitchell" 4/4 is feels sharper
Hey TeriyakiStyle, that last image looks really neat. Could you post a description of your workflow for the foliage? It's exactly what I've been looking for.
Did a little maxscript for both dimetric and isometric "squashed" (30° and 35.264) from OP mel script.
Then adjusting the camera FOV seems to be the easiest way to get the sprite size you want. It's kinda tricky to get the perfect 2.1 tiled shape tho.
--isometric
--settings for 118*60 sprites
(
--create cam
myCam = Freecamera pos:[180,-180,220] orthoProjection:true name:"isometricCam" FOV:74.05
m = myCam.transform
preRotate m (eulertoquat (eulerAngles 54.736 0 45))
myCam.transform = m
viewport.setCamera myCam
--create ref
myBox = Box lengthsegs:1 widthsegs:1 heightsegs:1 length:20 width:20 height:20 mapcoords:on pos:[-30,30,0] isSelected:on name:"refBox"
--set renderer
renderSceneDialog.close() --else doesn't register
renderWidth = 1024
renderHeight = 1024
renderPixelAspect = 0.8665 --this is where the squash happens
--no AA at all
--use "preferences > don't antialias against background" for no AA only on borders
scanlineRender.antiAliasing = false
scanlineRender.filterMaps = false
)
--dimetric
--settings for 118*60 sprites
(
--create cam
--had to offset on Z, to get right amount of pixel lines
myCam = Freecamera pos:[180,-180,220.2] orthoProjection:true name:"dimetricCam" FOV:74.05
m = myCam.transform
preRotate m (eulertoquat (eulerAngles 60 0 45))
myCam.transform = m
viewport.setCamera myCam
--create ref
myBox = Box lengthsegs:1 widthsegs:1 heightsegs:1 length:20 width:20 height:20 mapcoords:on pos:[-30,30,0] isSelected:on name:"refBox"
--set renderer
renderSceneDialog.close() --else doesn't register
renderWidth = 1024
renderHeight = 1024
renderPixelAspect = 1
--no AA at all
--use "preferences > don't antialias against background" for no AA only on borders
scanlineRender.antiAliasing = false
scanlineRender.filterMaps = false
)
and well, i'd used dimetric, and scale the cube if needed. Squashed iso seems a bit out of control.
I usually set up parallel projection with a perspective camera's FOV ( isometric )
Is the scale and rotation of camera primarily used for a predictable constant
( classic iso video game perspective and tiling? )
Otherwise is the isometric effect any different than a freely rotating perspective camera setup with parallel projection FOV? ( assuming one rotated to match your angle is tiling than not possible without the scaling? )
i'm not sure what you call a perspective cam with parallel projection (maya ?)
Assuming it's an axonometric cam, the type of projection just depends of angles.
For isometric, a cube will have the same length on the 3 axes.
Basically it's 45° on Y, 35.264° on X or Z.
People used to draw plans like this, because you can directly mesure any axes on the paper with a ruler and you get the correct size.
But most old games don't use iso but dimetric (only 2 axes mesuring the same), with a cam at 30°. Because you can fit naturally a 2/1 ratio and the tiles are perfect and match together with no pixel "holes". You can't make a square perfectly tile at pixel level with isometric projection, unless you squash it to get the 2/1 ratio like this:
i'm not sure what you call a perspective cam with parallel projection (maya ?)
Assuming it's an axonometric cam, the type of projection just depends of angles.
I always thought setting FOV as a Telecentric lens was orthographic parallel projection ( what makes a camera projection parallel is that any objects the same in size results in the same size/magnification at any distance? )
![[Deleted User]](https://polycount.com/themes/polycount/design/default-avatar.gif){kind=avatar}
Replies
Anyway, cool investigation and nice model you have there.
just my 2 cents
Can you explain the "Squish" a bit more ?
Then adjusting the camera FOV seems to be the easiest way to get the sprite size you want. It's kinda tricky to get the perfect 2.1 tiled shape tho.
--isometric --settings for 118*60 sprites ( --create cam myCam = Freecamera pos:[180,-180,220] orthoProjection:true name:"isometricCam" FOV:74.05 m = myCam.transform preRotate m (eulertoquat (eulerAngles 54.736 0 45)) myCam.transform = m viewport.setCamera myCam --create ref myBox = Box lengthsegs:1 widthsegs:1 heightsegs:1 length:20 width:20 height:20 mapcoords:on pos:[-30,30,0] isSelected:on name:"refBox" --set renderer renderSceneDialog.close() --else doesn't register renderWidth = 1024 renderHeight = 1024 renderPixelAspect = 0.8665 --this is where the squash happens --no AA at all --use "preferences > don't antialias against background" for no AA only on borders scanlineRender.antiAliasing = false scanlineRender.filterMaps = false ) --dimetric --settings for 118*60 sprites ( --create cam --had to offset on Z, to get right amount of pixel lines myCam = Freecamera pos:[180,-180,220.2] orthoProjection:true name:"dimetricCam" FOV:74.05 m = myCam.transform preRotate m (eulertoquat (eulerAngles 60 0 45)) myCam.transform = m viewport.setCamera myCam --create ref myBox = Box lengthsegs:1 widthsegs:1 heightsegs:1 length:20 width:20 height:20 mapcoords:on pos:[-30,30,0] isSelected:on name:"refBox" --set renderer renderSceneDialog.close() --else doesn't register renderWidth = 1024 renderHeight = 1024 renderPixelAspect = 1 --no AA at all --use "preferences > don't antialias against background" for no AA only on borders scanlineRender.antiAliasing = false scanlineRender.filterMaps = false )and well, i'd used dimetric, and scale the cube if needed. Squashed iso seems a bit out of control.
I usually set up parallel projection with a perspective camera's FOV ( isometric )
Is the scale and rotation of camera primarily used for a predictable constant
( classic iso video game perspective and tiling? )
Otherwise is the isometric effect any different than a freely rotating perspective camera setup with parallel projection FOV? ( assuming one rotated to match your angle is tiling than not possible without the scaling? )
Assuming it's an axonometric cam, the type of projection just depends of angles.
For isometric, a cube will have the same length on the 3 axes.
Basically it's 45° on Y, 35.264° on X or Z.
People used to draw plans like this, because you can directly mesure any axes on the paper with a ruler and you get the correct size.
But most old games don't use iso but dimetric (only 2 axes mesuring the same), with a cam at 30°. Because you can fit naturally a 2/1 ratio and the tiles are perfect and match together with no pixel "holes". You can't make a square perfectly tile at pixel level with isometric projection, unless you squash it to get the 2/1 ratio like this:
But see my example, squashed iso is a bit strange, not even considering animation. But yeah, then a cube could fit into a square texture.
https://en.wikipedia.org/wiki/Axonometric_projection#Four_types
https://en.wikipedia.org/wiki/Isometric_graphics_in_video_games_and_pixel_art
I always thought setting FOV as a Telecentric lens was orthographic parallel projection ( what makes a camera projection parallel is that any objects the same in size results in the same size/magnification at any distance? )
from https://en.wikipedia.org/wiki/Telecentric_lens#Object-space_telecentric_lenses