Pretty sure I'm gonna be shit on, but after reading through

>>2727064
The proportion part is talked in Scott Robertson's book and even Loomis's Fun with a Pencil. Basically just create a 2D floor plan.

>>2727064
Interiors where the horizon line is far off the image with lots of non-parallel objects that all have their own vanishing points can be easiest constructed by drawing a cube and using that as guideline. Then just eyeball it and make sure every object sits correctly on the ground plane. The side planes of the cube inform you over how the relative proportions change from the front to the back without needing accurate vanishing points.

Also, if you want to be more accurate, you can actually determine the exact horizon line in an image like this pretty easily. You just need to expand the canvas by a lot and then find 2 parallel lines that go towards the horizon line, like on the windows or the doors.

File: 1477637056082.png (723KB, 1563x735px) Image search: [Google]

723KB, 1563x735px

You can still determine the horizon line, it's just off the page.

Pic related, the thick red line is the horizon, determined by the converging of each wall of the room.
Purple lines show the perspective of a few objects.

Bright green lines show proportion of two students' chairs (imperfectly, since they're not perfectly parallel with each other - this could be worked out by drawing them parallel and rotating one with an ellipse in perspective).
Dark green lines show the proportion of two students' heads.

Best examples I could pick out in a complicated scene of overlapping objects and hazy lighting, hopefully makes sense.

But ditto to what the other two anons said. Scott Robertson's book can teach you a lot about these complex rules, and once you understand the theory behind how it works, it's very easy to just eyeball it.

Don't stress it, no one's going to become an expert on perspective after reading one book.

>>2727144
This anon's right.

What will really help you understand everything about perspective is Perspective from ctrlpaint.com shop.

File: 1427821802253.png (306KB, 593x540px) Image search: [Google]

306KB, 593x540px

>>2727064
>when I can't really determine the horizon line/ eye level in the first place?
you read that book and you can't even do that much?
try david chealsea's book.

>Like how would you know the relative proportion to everything
with a viewpoint like this you can almost just think of things in orthographic view, with the objects closer to you being only slightly larger than the ones in the back.

things that are small, and really far away (like this entire shot) don't necessarily need to have strong perspective, if you know what i mean.

maybe it's better to think about things in a more simple, orthographic way in order to get a grasp on size relations and distances.
beginners tend to get confused by the nature of converging lines, often they are too caught up with it to understand the more fundamental and practical usage of perspective...

i think it's a decent idea

>>2727144
>using extremely short lines in an already blurry picture to determine perspective.
don't fucking do this. almost all of those are wrong.
the red line on the left is completely off, you used THE worst line to extend. i'm not even sure it's a line, it's just a shadow.
you should've used or at least checked if the doorframe matches that VP. and it doesn't.

>>2727064

When you draw a realistic scene where either the horizon line or vanishing point is impractically far away it's better not to use orthogonal but instead use parallel lines, close to ortographic projection like >>2727223 says. You just use your eyes and a good understanding of perspective to eyeball it and only slightly vary the projection of lines from one end of the page to the next. When you get good at perspective visualization you don't need guidelines. I prefer this type of projection since it's more realistic and I always eyeball it. I've had people analyze my scenes and they end up correct

What you're essentially doing is one point perspective with objects turned away from the point of origin, so while size follows the same rules, objects don't align. When you use orthogonal lines it often turns out stiff and rigid, the world doesn't work like that and the FOV just isn't realistic

The other advantage is that you can construct natural scenes with altering elevation in a much less cumbersome way. It's important to keep in mind that the horizon line is not the actual horizon when your view angle is not at ground level

>>2727065
>>2727091
>>2727144
>>2727215
>>2727223
>>2727230
Thank you so much for the answers, I'll definitely be taking all of your advice when I study later

>>2727064
I'm trying to figure out how to go about lighting a scene like this thought greyscale + overlay of the local color would work but it looks like shit

You've just gotta feel it man. No really, the pros don't think about all the stuff they talk about in perspective made easy, it's all internalised. If you over analyze perspective and draw a million vanishing points it's going to kill your work. Try to eyeball things more

>>2727403
This.
If you look very, very closely, the perspective of >>2727064 isn't flawless (the angle of the wall with the door gives it away), but it works anyway.

>>2727223
>almost all of those are wrong.
kek, that's what I get for fucking around on /ic/ instead of going to bed, dunno what I the hell was thinking on the left side.

This dude's right, OP, but hopefully the basic idea at least makes some sense. Like it said, it's a tricky image to retroactively work out perspective on, and obviously I wasn't being very diligent.