Precise control of camera coordinates and parameters in SketchUp

[ViggoPaulman]

Hello,

Enscape has this very neat feature where it allows you to edit camera coordinates for each scene. This is very handy when you need precise camera placement - achieving perfect symmetry, precise tilt angle, exact distance from an object etc (see screenshot below):

I’m looking for exactly the same functionality but inside SketchUp itself. Is anyone aware of a plugin where I can precisely input coordinate values of the camera in relation to the world axis and adjust tilt and yaw angles? I found this thread from 2018 but the links shared by George no longer work.

[franciscofiallosb]

I use fredo portrait to create axonometric views at different angles, the default isometric view of sketchup is 30/60. If you want to place the camera to a certain distance you can use the native place camera tool just draw a line or something to snap the camera. It’s also possible to create cameras with different lenses using the advanced camera tools plugin.

[ViggoPaulman]

Hey Francisco, thank you for the reply.

I’m trying to achieve something different. I have a huge facade (over 7 meters in width and about 5 meters high) that needs to be fully rendered and printed at 1:1 scale.
No rendering engine can produce a single picture with enough high resolution for that. The only way to achieve this is to place multiple cameras at precise intervals/coordinates, batch render and stitch everything together in post.

That’s why I need a coordinate/position input system.
I could achieve this in Enscape, but the issue is my SketchUp file is in imperial and Enscape only shows units in metric.

[Box]

@ViggoPaulman

the issue is my SketchUp file is in imperial and Enscape only shows units in metric.

Change the units in sketchup.

[ViggoPaulman]

Converting and tracking precise coordinate decimal fractions of over 4 dozen cameras is not an optimal way to approach this, I’ve already thought of that. Also, parallel projection mode in Enscape and SketchUp are not well optimised, so inputting the coordinate values in Enscape is sadly not going to bring me anything. This needs to be done in SketchUp, hence my question - if anyone is aware of a plugin for SU.

[franciscofiallosb]

Use layout, set the viewports to 1:1, chose the biggest paper you can be able to print and place the part of the facade on the page, then copy the viewport on a new page, actually two viewports the first one for reference and move the handles of the second so it matches the continuation of the first one, do it until you have all the pieces ready to print, export to pdf and send them to the plotter.

[RTCool]

You could draw a grid or other physical geometry - whatever you need - for where to put the camera, and then use the position camera tool place and point the camera by snapping to end points in that geometry. (You can’t control yaw however.)

There is also Advanced Camera Tools, but it sounds like you’re only interested in orthographic projections here, not perspective views?

[bmike]

@ViggoPaulman

I’m trying to achieve something different. I have a huge facade (over 7 meters in width and about 5 meters high) that needs to be fully rendered and printed at 1:1 scale.
No rendering engine can produce a single picture with enough high resolution for that.

I assume you need something with materials and shadows, so you need a raster image as final output?

What is your target resolution for the final print? Have you taken viewing distance into account when thinking about final output resolution? Do you need 300dpi for someone’s nose to be pressed to the image?

From:

Photography Tutorial - Resolution, a matter of distance

VRay has a max number of pixels you can render:

Please keep in mind that the total number of pixels in V-Ray Frame Buffer is currently limited to [2 147 483 647](tel:2 147 483 647) pixels. For a square image, this means a maximum of around 46340 х 46340 pixels; for a width/height aspect of 2:1, around 65 500 x 32 750 pixels.

At 150 pixels per inch (which would be a very close viewing distance for a print this large) you need a long edge of 41,400 pixels. Your image resolution would be 41,400x29,400=1’217’160’000 - which you could theoretically do in VRay.

[Cyentruk]

You can simply use Position Camera (1) like this: click and hold your left mouse button at the precise position (2) then drag to the direction (3).

[bmike]

@Cyentruk OP is trying to move the camera in a grid across the facade while keeping the position perfectly aligned and is looking for techniques on how to achieve this.

[bmike]

@ViggoPaulman

The only way to achieve this is to place multiple cameras at precise intervals/coordinates, batch render and stitch everything together in post.

I think this isn’t quite true. I stitch panoramas all the time using Lightroom, Photoshop or PTgui. You would need to be in ortho view, aligned to the surface you are rendering, and the same distance from the subject. With some overlap in the tiles you don’t need perfect grid position of the camera.

I would reverse your thinking on this - if the facade is a group or component you could keep the camera in one position and move the facade. This way it is easy to keep the camera properties exactly the same. And if you have enough overlap between images I would assume Photoshop or similar would be able to align and stitch without too much trouble. This assumes that you are using a light source far enough away from the facade where this type of movement would yield realistic looking shadows in the final output after the movement, which with a global sun or soft box setup seems reasonable to assume.

An alternate would be to make copies of your facade moved into the correct position and then use one static camera position and create new scenes that capture the facade copies that have been moved into the correct position and assigned appropriate tags.

[panixia]

@bmike

which you could theoretically do in VRay

As long as you have theoretically infinite RAM and theoretically infinite time to devote to waiting for such monstrosities to render.
You’d be better off simply sending this project to a render farm.

[bmike]

I send all my 5k final renders to Chaos Cloud. Only local renders are for setup, test, etc. I assumed something this large would have to go there…

[ViggoPaulman]

Thank you everyone for your helpful comments and input.

Yes that’s true, my aim is to move the camera in a grid-like pattern, perfectly positioned and perpendicular to the facade and make renderings, which I would then stitch together in Photoshop to get one master high-resolution print file.

@bmike to answer some of the questions:
Yes, I did take into consideration the viewing angle. 300 ppi would be overkill for this print as you rightly say. After some tests I chose 127 ppi as absolute minimum, to ensure sharpness. My initial calculations tell me the horizontal width of the master print file should be 36,576 pixels. If I place 6 cameras on a horizontal run, then each render should be 6096 pixels, which should be a manageable thing to achieve.

Moving the whole architecture while keeping the camera static was a genius idea, thank you!

Yesterday I tried capturing one horizontal line of cameras with overlaps to stitch them in Photoshop as a test. However, it seems Photoshop inconsistent with stitching, either that or I’m doing something wrong:

After importing and arranging the renders (you can notice each frame contains an overlap of the other), I tried Edit > Auto-Align Layers > Reposition:

I end up with this mess:

or this:

The strange thing is, after trying the same a couple of times, it did align/overlap them perfectly. But the results are inconsistent and cannot be replicated.

---

If I could make renders that don’t contain overlaps, this would be a much easier job, but I understand it’s impossible to calculate precise interval camera movement when you’re rendering in Parallel Projection.

I’ll do some more tests today. If nothing works then I will have to manually align and stitch them in Photoshop. Thank you!

[Anssi]

@bmike

At 150 pixels per inch (which would be a very close viewing distance for a print this large)

When historical buildings at prominent locations are undergoing repairs, the scaffolding is sometimes covered with a canvas image depicting the original facade at full scale. The resolution to print these large posters is typically much lower than that, 36 dpi is often used. A building-sized image would probably work with an even lower resolution.

[bmike]

Glad I can help.
Years ago I managed to stitch some macro camera work (sort if like a copy stand setup_ as a flat image / page… but I just tried Ps and get the same result as you. I also tried PTgui and it doesn’t work without manually placing the control points. There must be a software solution - but I haven’t found it. Hugin or AutoPano Giga might also work.

But in Ps this is relative easy with your careful camera setup - but it is time consuming to do manually. To keep things easy I would manage your rows and then combine the rows and not try to do it all in 1 Ps document.

Import a row of images as layers, then make the width of the canvas the total pixel width you need. From there you can change the layer style to ‘Difference’ - sliding each image to the left / right and aligning the pixels is relatively easy when you can see the difference of the layer above. Do this for each image toggling on and off the ‘difference’ setting

I would then do a ‘save as’, flatten the image, and when you have all the rows you can repeat to stack the rows on each other.

Here is a crude example done just with random screen shots of a project currently open in SKP. You will have a much easier time as you will only need to slide the images left / right to align, or up / down once you have the rows built, and your output pixel ratio will be the same across all images - so no need to scale to align.

When things are aligned it looks like this:

Finished:

[bmike]

I agree - I’ve done some big photo prints that were setup for 125dpi, but when I worked in marketing / advertising I was astonished that billboards were done at 48 or less, depending on the size.

[franciscofiallosb]

If you render the viewports as vectors or hybrid you’ll have the best resolution, it will take some time to export though. I don’t think it’s worth it using vectors on something that big, just raster using high or medium quality would be enough and will save you a lot of time. I haven’t printed in scale 1:1 in a long time but the last time I did, was using the method I mentioned above. the width of a plotter paper is A0, 1.19m and the length can be up to the length of an entire paper roll. You could create your own paper size on layout. You would need between 10 or 15 pages depending on the length you decide to create the custom paper.

[numerobis]

I’ve build a small plugin to generate ortho cameras based on the current window aspect ratio and a selected facade (with help from my personal coder Claude ) Maybe it works for you…

facade_camera_generator.rbz (14.8 KB)

[ViggoPaulman]

Hey there @numerobis, thank you, sounds interesting! Do you have a YouTube tutorial or a manual on how to use it?