From web dev to 3D: Learning 3D modeling
levels.io
levels.io
I know very talented people and i guess that unless land a well paying job in the gaming or movie industry (good luck with that) you end up freelancing for construction companies and architects. Aligning sofas and tables...
I want to do stuff with VR though as a maker, entrepreneur, artist, or however you'd like to call it. If it's monetizable that's a fun side to it. But that's not the priority here. Fun is!
I do see it as a skill in the future, just like coding is now. If you can combine that with making an app, game, art, business or anything really, that's good for me.
That usually scares off all but the most serious funding.
Once you start playing with Unity you'll see there's a whole nother side of the 3D pipeline that takes just as much effort.
There's a reason that games budgets end up running very high unless you're dealing with 2D/simplified content.
For any given asset you may have a pipeline like so:
Sculpt hi rez model in Z brush > create low rez model and retopologize (can be a very manual process) > UV unwrap and bake the high rez details down to Normal maps for the low rez mesh > Use DDO or Substance painter to create custom textures for diffuse, metalness, etc > Rig model (if you are animating it) Then export to the engine. Rinse and repeat based on major art changes.
If you're already in Cinema 4D you should look into the various post-processing and rendering options. I think a little bit of DoF, ambient occlusion, global illumination and volumetric light could really make your scene shine.
Pros and amateurs use drawing programs quite differently. Pros don't edit much. They redraw. They don't spend a lot of time pulling control points around.
There are 3,467 union animators in Hollywood, according to The Animation Guild. It's not a big field.
Simply providing a clipping path which separates the foreground from a background is widely outsourced. Effects movies use this heavily. It's manual green screen. Some outsourcing companies for this job:
clippingpathindia.com
offshoreclippingpath.com
clippingpathasia.com
That being said the state of the industry in Hollywood, especially for animators and artists, is pretty dire.
When I flew to San Francisco for the first time, I was shocked that I recognized the all bay scenery from having played many hours of Flight Simulator going from Oakland to SFO.
The public pressure helps to keep me accountable, and it's more fun if I have an audience :)
- You need considerable 3D domain knowledge to employ the scanning tool correctly in the first place, and to post-process it's results
- The scanning tools don't work in many situations - have fun scanning something transparent, reflective, that won't sit still etc etc etc. The curtains in this scene, or the translucent lamps, and so on would all be very, very problematic
- The scan gives a noisier result than doing it by hand (Look at the model quality in figure 1, last step - it's pretty bad - that would be really, really apparent in a higher-res moving scene). High-end laser or contact scanners are better, but very expensive.
- Scan topography is currently really bad from every suite I've seen, and not really ideal for anything other 3D printing or using as a reference for building a proper 3D model.
That's too broad to be true and your examples below don't say anything about the fundamentals of 3D scanning.
It also ignores causality - in that some things are built in 3D first and then fabricated (extruded/milled etc...) while others are fabricated then scanned. So the 3D data around each can't really be compared in the same sense.
In fact photogrammetry and laser scanning is getting to the point of eclipsing hand modeling for pretty much anything that is not a result of a CAD model.
See: Our photogrammetry patent
The point of my examples was that scanning is currently so limited as to make a pure scanning workflow unworkable in almost all situations, including this one. Scanning is incapable of doing what modelling can (good topo, clean results, accurate representations of translucent, reflective, etc surfaces). Likewise, modelling is incapable of doing what scanning can (quick approximate real-world captures). They're different tools that accomplish different things.
> In fact photogrammetry and laser scanning is getting to the point of eclipsing hand modeling for pretty much anything that is not a result of a CAD model.
Due to differing definitions of what is considered "CAD" I'm not sure what you mean by this. I can think of applications where 3D scans are used, but only either as a reference for hand-modelling, or only after being so heavily hand corrected that it's a really more of a hybrid (model/scan) process.
Animation almost always uses hand-modelled or heavily corrected (by hand) scans. Manufacturing uses hand-modelled objects or traced-over scans. Gaming uses hand-modelled objects or traced-over scans. Where are scans eclipsing hand 3D modelling? They're being used but as a different part of a hybrid workflow.
Going further, with already built things, that aren't previously built in CAD, like components and old buildings, in solidworks autocad etc... Making those models is much more efficient with structured light or photogrammetry systems. It should be noted that it's not just speed, it's hard to model to micron specificity, whereas we can with scanning.
Animation is a little different but I think follows the same principles in the long run.
The distinction between the methods in my mind is between digitizing something real versus making something new. I think in the former case scanning will win long term.
> Making those models is much more efficient with structured light or photogrammetry systems
Only if you don't care about topology, representing unscannable things, polygon efficiency, UV layout, rigging, etc etc etc. For animation it's nowhere near competitive at present - you basically have to resculpt or retopo everything for anything but static matte background props.
Maybe long-run pure scanning workflows will become more competitive, but that's not what my comment was about.
Not true. In fact those are important aspects - especially UV layout as texture baking correctly across dynamic finishes with the same underlying polygons is critical for accurate rendering.
majority of the various 3D industries
Like which? Aside from production level CAD/CAM we cover most other use cases as they are almost strictly rendering/modeling.
Only if you don't care about topology, representing unscannable things, polygon efficiency, UV layout, rigging
Again, not true. Especially polygon efficiency - our model requires polygon optimization so that our models can be utilized by low-bandwidth areas. It has been a massive challenge but one we cracked big nuts on. "Unscannable things" is an issue, but again, there are solutions including destructive scanning and X-Ray/MRI which give us insight.
Woah, interesting. Please go on? (i.e. what would be your workflow for this - what applications/apps/hardware would you use for this? Also, for a thing with a lot of polygons, like a human, ok -- but would you do same for an object better suited for CAD programs? Like a monitor, let's say?)
As I mentioned elsewhere the core distinction is in digitizing real things that did not come out of a manufacturing process that utilizes CAD.
For us, as we are an AR company, building 3D content is a critical piece so we need to digitize real objects as quickly and accurately as possible.
The trick for us is finding the perfect balance between size of the resulting model and quality. So our scanning process is very refined.
Our product is real-time and "other algorithmic tools" are the basis for our work - and we have had fantastic unprecedented results.
Raw scan data is a fantastic starting point but it's not an end to end solution.
That is true for current COTS tools, but what we built has proven that we can get an end-to-end solutions for our problem which is opaque, non-reflective objects larger than a soda can and smaller than a vehicle.
I know everyone on HN is the smartest ever and nobody here can see other people as making real shit but there are innovations happening. In this case the wide scale need wasn't there for innovation for a long time and now there is.
But "a picture of a naked person" is as much the goal of figure drawing as I suspect "a model of this street" was of this project; the real goal is the knowledge it puts in your head, both of your tools and of your subject, which you can then use in making stuff up from scratch.
I mean, just texturing and lighting alone makes it a huge deal as opposed to its competitors. It's simple and makes everything look beautiful. Texturing is drag 'n drop and you have access to global lighting (which doesn't exist in Blender from the get-go).
---
Granted I haven't touched Cinema4D since 2003 :) it may have changed a lot.
I started with 3ds r4, then 3ds Max r2, blender, lightwave, maya. I've stuck with blender the most, partially because the ui is fast to use for me. It was not fast to learn, though that has improved.
Granted, Cycles (Blender's global illumination renderer) is not selected by default, but it's a drop down at the top menu bar to switch.
Then you can click "file" -> save "default settings".
Here's the official Cycles demo reel: https://m.youtube.com/watch?v=wDRTjzLNK0g
A few things:
- Non organic modeling is very programmatic, seek for symmetry in all things. Train yourself on wheel rims for instance. Then bevel, chamfer. All these details are very 'procedural' but they make the difference between a PS2 feeling and a feeling of beauty. Be lazy in a smart way, replicate and randomize the secondary details to give density to a scene. I don't know what blender gives today, try other tools, zbrush, modo, houdini, maya. They'll give you ways to express your ideas.
- Organic modeling... Andrew Loomis books about human figure gave a few life altering tips (most of them are free PDFs now). Also, search for mirai modeling bay raitt, martin kroll on youtube.
- Lightning today global PR based lightning does a lot for you, but scenes are about 'emotion' not physics. Peek what you show, what you don't, and how you do. Hollywood movies are known reference for that, they exagerate, emphasize everything to draw attention and decorate important points in the composition. Unless your goal is to convey a more natural sensation in which case forget everything about hollywood :)
Enjoy
For years afterward, I confused Imagine with Lightwave, but of course they aren't actually related.
I just wrote a blog post about the same thing as it pertains to breaking into VR in general. Interestingly, I had dedicated a chunk to the prerequisite of 3D modeling and animation.
Mine is no where near as good a post since I'm not an experienced blogger by any stretch of the imagination. But complementary I suppose..
I posted the link to HN here: https://news.ycombinator.com/item?id=11733068
If you can export it as obj/mtl then you can probably just adapt this code: http://threejs.org/examples/#webgl_loader_obj_mtl
So I would consider Sketchup as the Ruby of 3D.
Sketchup is underrated and considered 3D for amateurs but people don't know that most architectural offices use it during early stages of projects. You can't beat its speed and usability.
For example, if I want to model a coffee cup, I might define the cross section then revolve it 360 degrees. I know of OpenCascade, but was wondering if there were others.
http://iquilezles.org/www/articles/distfunctions/distfunctio...
Don't understand why you'd want to so the coffee cup example programatically though, can be done nicely parametric from the GUI, with more direct manipulation..
The best part of this article is when a learning endeavor turned into a visit to another country.
I just wish I knew how to make pretty things with it...
I did not design the model[0], but asked the artist to export it after the animated video was done and used it to make a simple gamified online shrine. It was a lot of fun to reason in 3D for a change.
Is there a simple transition from this program into Unity?