Home Artists Posts Import Register

Content

I guess the title shouldn't necessarily be "using fSpy", since the fSpy part is only like 4% of the total process, and the rest is in Blender, but it's a great way to get the camera aligned to a photo, and makes the rest of the process way easier, since once you have it lined up, you can just extrude faces out along the various axis!

If you have an image of a building (or an interior), and aren't sure how to start modeling it, fSpy is just a really great way to get started.

You can download the application here, and there's a link to the blender addon on the same page :D

Here's the grocery store picture of my local Albertsons if you want to give it a shot. 

Files

Creating Grocery Store/Rando-Building with F-Spy

Comments

Anonymous

Maybe that is actually the best tutorial about fspy around! Thanks a lot, that actually helps!

Anonymous

Have you ever tried fpying two or more images taken from different perspectives, placing the cameras correctly in 3D space and using both images for projection? I gave it a try and I couldn't get it to work well, fairly new to Blender though!

IanHubert

Ooo- I haven't, no. Dang yeah that does sound tricky, though. In the grocery store example, the only real point-of-reference depth-wise was where the object intersected the floor, and that gives a ton of wiggle room for stuff to get out-of-alignment. That said, having a couple extra images from other angles could work really well for recreating all the textures! And add enough more and you could just do a photoscan, hahaha

Anonymous

I'm going to explore it some more. One method I tried rather than using fspy was to create an image sequence from each of the images, motion track and then solve, I need to go back to it but it seemed like it could actually work. I'll let you know if I discover anything that could be of use!

Anonymous

I’ve experimented a tiny bit with doing a low detail (quick, 10 phone pics) photoscan, then projecting on the photoscanned mesh. That way you don’t have to model every detail but still get a nicely detailed image (from a limited range of angles).

Anonymous

Hey Ian....thanks I do enjoy these...hey, have you ever modelled a pinball machine? (just throwing a random question out there to give you ideas for another post ;) ;)

Anonymous

there has to be a way because thats how keentools facebuilder works. You use multiple images to create a mesh

Anonymous

in alot of your videos you always turn on clip instead of repeat but in this noooo

Anonymous

Question: how do you deal with copyright stuff? Like if this was a paid project would you paint out and replace the logos? Or just put them behind a grungy window?

Kai Christensen

Posted at 3:30 AM... working another deceptive freetime unit, are we?

Anonymous

Great tutorial, Ian, I love it. Such amazing results in so little time. Has a lot to do with your expertise in Blender, I still feel I'm playing back in slow motion. I'm still puzzled why nearly nobody uses quick favourites, I couldn't live with my quick-color-ramp. :^{O

Anonymous

I often wonder the same thing. If a logo is blurry enough, I doubt any company would care. Sure, everyone would notice that those 3 pixels looks like Heineken, but what if you replaced all the logos with a parody logo of "Hindcan"? I assume the general rule of thumb, is if you can clearly read the logo and it's in focus (implying it could be a paid spot/endorsement), then you may need to get permission. If you're just playing around and making stuff, I don't think you should worry about it.

Anonymous

Soooo this is amazing I didn't realize there was a different quick favorites for shaders... just added colorramp to my quick favorites hahaha thank you for this!

Anonymous

Yeah, it's contextual, the favourites you will see will be different for edit mode, object mode, shading, compositing... Pretty handy stuff 😊

Anonymous

dude how much ram do you have??!! How does your computer even work with that many chrome tabs!!?? XD

Kai Christensen

Or with that much geometry in edit mode? Like I know it's not even that much, but my system lags once I get to like 1000 vertices...

Anonymous

I mean, if you think about it, all you're really doing here is using the classic front-side-top image reference views like from every modeling tutorial in 2008, but with arbitrary angles! So adding more photos is totally reasonable. Facebuilder's cool trick is solving the camera positions/focal lengths and modeling the head at the same time, which definitely involves some fancy tricks—that's also why it's more vulnerable to bad results, since the back-and-forth between camera and object solving can get into a feedback loop where they both ruin each other.

Anonymous

Texture-wise, you can mimic what Facebuilder or any photoscanning software does by mixing your image textures using only the best view of each surface (okay, it's generous to say photoscanning software always uses the "best" view of each surface, I mean, that's where blurry or discolored texture patches come from, but that's what user intervention is for). The appeal of that is, among other things, fixing occlusion issues like the shelves or pillar in this video. If you want to mix textures like that, I’d say the simplest way is just having a material for each angle, with the correct image loaded in, and then any errant faces can simply be reassigned and reprojected from the correct angle.

Anonymous

That’d probably work best for boxy objects like these shelves or buildings—if you want to actually blend your textures, that’s fairly easy too—check out Jan van den Hemel’s recent Blender Secrets video on texture projection painting (Jul 21—I’m not gonna add the link cause my comments with non-Patreon links in them always seem to disappear, I think Patreon doesn’t like comment links for some reason?) That lets you actually blend the influence of your different angle textures across your mesh with feathering and opacity.

IanHubert

Oooooo- you're totally right! I should get into Quick Favorites! Especially for nodes, that'd be fantastic.

IanHubert

Yeah honnnnnestly I feel like even the "Official Laws" are a little fuzzy on this. Coke pays millions to plaster their logo everywhere- can they really sue you if their logo is incidentally in the background of a shot? And even if they can, HAVE they ever? I've known art teams that've made entirely aisles of fake products, and I know other productions that've just filmed in the store as-is, and I dunno, unless if you're SAYING something about the brand, I don't think Quaker Oats is going to come after you because they're in the background of a shot. What I'd typically do in a case like this, if this were for a paid project, is just go in in photoshop and screw up each logo a little bit, kind of like how Abraham is talking; screw up any particularly distinctive logo. HOPEFULLY not having to generate new ones... That said, even on huge projects I've worked on with major artists/studios, the general rule of thumb is, "EHHHHHHH I dunno I guess we should paint it out to be safe why not?" (And I'm sitting there like "oh oh the 4 days of human life spent rotoscoping is why not" buuuut that's the gig.) Anyways yeah, I feel like people are usually quick to point out that "Fair Use" comes into play for criticisms, but there does seem to be a good "don't worry about it" element.

IanHubert

Ahahaha, allllwaaaayyyyyys. Sean and Kaitlin are home all the time now because of Covid, which is nice, but it also means I don't really get that hyperfocus time till folks go to bed at midnight. :D

Anonymous

Dis boi wild

Anonymous

COOL IAN ! I LOVE IT WHEN YOU MAKE SOMETHING OUT OF NOTHING ! BUTTTTTTTTT .....

Anonymous

brilliant as always mate

IanHubert

Oooooo- that'd be fun! Especially if we did some stuff with physics so it was actually awkwardly playable, hahaha

Anonymous

random human pinball using the characters you scanned... weird...

Anonymous

Love the tutorial and - to be honest - also love the picture of the store. Because holidays this year will at home, (covid) so I look at pictures of "normal" places over the world for some "imaginary" road trips. (I usually visit "normal" stores in foreign countries.) So double thanks

Anonymous

The real reason for blurred labels or generic stand-ins is all about advertising, cross promotion and product placement revenue. Let's say the leading lady in your movie or TV show is drinking a Coke because that's what you happened to have on hand. Well, now you've a) given Coke a free product placement (thus throwing away potential $$$) and b) have eliminated any chance that PepsiCo or any of its brands (Doritos, Cheetos, Gatorade, etc) will buy any advertising relating to your project, let alone partner with you on a promotion deal.

Anonymous

I've seen lots of people asking if you can just put these into a game engine for VR, so I tried it out! (you totally can.) https://youtu.be/AWKrEoiELsQ

Anonymous

Hello Ian, what do You think about quickshape v2 from Jama Jurabaev https://www.youtube.com/watch?v=oEkSq6yqGs0 ? seems much faster than the knife tool.

Anonymous

It honestly boggles my mind how messy you work but still get the most realistic results. I took an animation class with digital content creation in high school and currently taking another one in my freshman year of college and watching you just copy faces and place them on top of each other while overlapping makes me go insane. I need your wisdom on this...should I continue "good practices" or should I start to be messy. I think I can kind of answer my own question by only being messy on things that will add to the world but won't really be focused on, much like the grocery store, and then focusing good topography and texturing with focal objects. But I still want to hear what you have to say. This is my first time ever participating on Patreon and so far I've learned so much from your Patreon and I'm loving it. Thanks for the inspiration and honestly its incredible. Also I'm sorry for how long this message is lol.