BTW Using decals in SketchUp is pretty cool:

Whoa. Great catch Alex - its the Vector3d#axes method I was missing.

Be great to have the code behind the method (pretty please?) - the problem is that while you can generate orthogonal vectors from a normal easily enough, there is some ambiguity regarding orientation; this method obviously captures how Google chose to resolve it - but be good to know what rule they chose.

LightUp too can now handle Images - they're great for adding surface details to models. Water stains, scratches, rubbish etc.

[And thankyou Thomthom for the link.]

Adam

Anyone had any success extracting the local transform for a SketchUp::Image?

@mike lucey said:

@solo said:

I don't get the 'boilerplate component' reference, care to explain?

I think he may mean standard or common place. Yep, some AutoCAD files I've come across are very 'leaky' indeed.

Yes, thats it. Often times you have Primary assets which are done by yourself but you may populate/situate with "off the shelve" stuff which has little to do with what you're trying to do but is none the less necessary.

The one that motivated my post was a model of an (artistic) installation built on raised floor panels. The floor panelling is kinda irrelevant but was 200,000 faces of crap.

So you can understand people not wanting to burn cycles on this peripheral stuff but then getting let down by the poor quality.

Its been discussed before, but perhaps a tool that shows relative geometry complexity might be good - I've certainly had entire office floors with less geometry than in a single desk chair in the corner of the building..

Adam

I keep coming across SketchUp models that use boilerplate Components that appear to have been imported using DWG that are uniformly terrible quality - with cylinders particularly having loads of long skinny shard triangles that don't abut correctly, have gaps, wrong smoothing groups etc.

And/or are massively over the top wrt detail. Coat pegs with N thousand faces, cross bracing where each element is thousands of faces each. Mad and makes for enormous SU files.

What gives? In the video games biz, we have hard geometry budgets that are required to be met for each component as well as large automated conditioning pipelines that all geometry flows through to ensure that it is defect free.

So are there sources of SU components that tend to be better than others on this front? And this big variability in quality; is it normal in Arch / previz?

Adam

@unknownuser said:

I have often wondered how a jpg with no compression loses data when it is saved. I am not sure that is the case. The fine pixel below has a set of original pixels on the left and the same set on the right after ten saves, both from a much larger image. Each time the new image has been saved, opened, then re-saved. This is not an elaborate test but as you can see if there is data loss I don't see it. It is my guess that the application saving the data is leaking and not the jpg itself.

Its a bit of an unfortunate choice of image because JPEG compression works by breaking out chroma and luma and storing less resolution for chroma than luma. (I appreciate concrete is mostly grey). Hence you will see color shifting of pixels as JPEG compression increases - typically grays become a little pink for example.

JPEG is by definition a lossy format. Formats such as BMP,TGA,PNG are lossless.

However, it is also true that if the color gamut of the image you're saving is sufficiently small, you'll get almost no loss using JPEG at highest quality. So a largely grey concrete texture may well be fine. A photo with a larger range of colors absolutely will show loss using JPEG. Nothing to do with dodgy Apps losing data.

If you're concerned about loss of quality and you have the disk space, there is no question that you should use a lossless format that stores the exact pixel values one after another in a file. But if you're dealing with a 4000x4000 pixel image, thats 64MB of data. I'd guess JPEG - with neglible quality loss - would be able to store it in 10% of that space.

Adam

[Great video Ecuadorian!]

@jsteacy said:

Very cool,
I have some questions about lightup, I thinking about buying it. Did lightup create the video file for you or did it have to process a bunch of images or use any other software? Did it take a while to generate the video? How big is the video file?

LightUp has a "Record movie" function which works in realtime. You start recording and can use preset Scenes and/or just fly/walk around the model recording your movie. Its then saved to disk as an uncompressed AVI file. The uncompressed part is good in that you get lossless quality which you can take into other packages. However, the AVI files are large so if you want to share with clients, you'll need to pass it through a codec to crunch it down (eg MPEG-4 / H.264)

Alternatively, you can export the entire model with lighting to a custom file for viewing using the free LightUp Player.

Just got bitten by this in LightUp, so be warned..

On a Mac, filepaths using widechar to accommodate Chinese/Japanese etc work fine. Unfortunately, Ruby 1.8.x on Windows (ie SketchUp Ruby) uses ANSI file handling, so you get filenames like "?????" from Dir.entries that include Chinese etc filenames/folders. Ditto FileTest.xxx functionality.

I've had to write a bunch of widechar file handling extensions this morning to fix LightUp file handling on multi-lingual filesystems. (Be in LightUp v1.9c)

Adam

Sure you know this, but if you're comparing distances rather than interested in the absolute distance, the fastest way will be to project a vector onto itself using scalar product.

p1.dot(p1) < p2.dot(p2)

Adam

I think I can see what is happening. Revit is trying to coalesce triangles. You can see this from your pic.
I think what is happening is the tolerances for flatness are being confused the scale of the curvature. Imagine a single triangle. Is has 3 edge neighbours which are potential candidate for merging. If the curve is sufficiently small scale then Revit could mistaken choose a neighbour on the curve rather than the triangle that completes a Quad ..again and again.

So see if there is a tolerance for flatness/planarity that can be tweaked. You could also try doing a scale up of your model in Revit before export to workaround any hardwired tolerances in the exporter.

Adam

It would be trivial to write a Ruby that welded coplanar triangles back into polygons.

Or is this something different?

Adam

you sure, TIG?

I'm certain I tried doing a html+js progress bar and found its actually run from inside the Ruby execution loop.

Whoa there! Hammer to crack a nut there.

In a right handed system (like SU), the front of a Face is defined in an anti-clockwise direction. Imagine gripping the face normal with your right hand and your thumb in the direction of the normal - your fingers show the winding of the face.

Face#pushpull() takes a distance to move along the face normal. So if you define a Face and paint its front material with something, then apply a negative pushpull, you'll always get a cuboid with that material on the outside.

Adam

Nice thomthom, but wouldn't it be simpler to apply color to the face and then pushpull and it will pick up the material of the originating face.

OK, in video games running under Windows, we always go for the former. ie Keep it hidden and centered and generate deltas. potentially getting the root window handle's mouse events also (ie the whole desktop).

However, in the context of a Windowed Application, you might consider a different metaphor. Perhaps look at how GoogleMaps uses a "virtual trackball" in the corner of the display?

Adam

Pixero, can you share what you're trying to do?

Its generally viewed as bad practice for programs to move mouse cursors rather than allow the user control of interactions. Is there another way of expressing what you want?

Adam

Mousewheel is currently not supported. We'll look into it.

@thomthom said:

One thing I struggled with when navigation was to get the camera to stop. The inertia effect keep drifting the camera further away than I want.

The idea behind LU Player is to capture the setup the author chose using LightUp. So if you create the content to have a camera speed of X and no inertia and using a specific Skybox, this is what you'll get in LU Player. So if you don't want inertia, then export with inertia unchecked.

@gaieus said:

Also, would there be a chance to "add speed"? (In some FPS style game engines you toggle a "running" mode with the shift key). With some bigger models this speed would be a bit slow if you want to wander all around...

Hold Shift down before pressing direction keys for 5x speed increase. Also using Scenes (right mouse click) to get to a vicinity is handy too.

@thomthom said:

Is there any way to enable SketchUp like navigation in LightUp Player?

What do you have in mind? Always interested in getting feedback.
Currently, in LU Player, you can pivot around a point (Alt/Ctrl) as well as navigate around your model.

Gaieus: LightUp Player has exactly the same features and behavior as LightUp inside SketchUp.

So, yes. Any animated Dynamic Components setup in LightUp, texture animation and color animation as well as SketchUp Scenes work inside LightUp Player. It also can run Fullscreen for presentations etc.

dale. The light rigging + shader parameters should be able to be exported from LightUp. The texture animation also but in that case there isn't other packages that support that kind of thing.

FYI The webpage for LightUp Player is here http://www.light-up.co.uk/player

Adam