Hey everyone. I’m working on a new business providing (nearly) perspective-free orthographic reference photos of people.
I recently did a proof-of-concept photo shoot to test my method of getting an orthographic image of a person, and here are the results. Notice how you’re looking straight on from the head all the way down to the feet vs. (in a typical reference photo) straight on with the head, but eventually the tops of the feet with varying degrees of perspective depending on how long of a lens was used.
These are lo-res 512px versions, but links to higher-res 2k samples of the files are hosted here: reference.silentNstudios.com. Because these are just samples from a proof-of-concept test shoot, these are the only views that are currently available, but I’d love to get your feedback as artists for what else you need from reference images.
What I currently plan on offering are:
Better accuracy: Nearly perspective-free orthographic reference photos of people (beyond just stepping farther back and shooting with a long lens)
Less guesswork: The models will be shot with an emphasis on consistent posing between different views so that everything lines up from front, side, back, etc. (No head tilts, shifted stance, etc. between views)
Save time: The images will be pre-aligned and imported as reference planes in the file format of your choice (Maya and Blender to start, more formats to follow) so they are ready to start modeling with immediately.
Better textures: Models will be lit with uniform, flat lighting for better texture information. Specular highlights will be minimized or eliminated from dark-skinned models, and shadows will be equalized on light-skinned models. Full body images will be available in ~5k x ~5k resolution.
As I said, I’m just starting, so I’d love to get feedback for what you need while I’m still in the early stages of this business.
I am a 3d.sk paying customer and often was annoyed that the images are shot with a lot of perspective or targeted upwards or downwards resulting in a lot of work using them as modelling reference, generally tough to match up 2 views and usually resulting in a person not really looking like the reference unless you shaped the body afterwards with a perspective camera, which then was tough again because the FOV was unknown as many images lacked a proper EXIF.
It’s not necessarily bad to have a new version of the person but sometimes I wished to have exactly the reference look just by working after the images, not by tweaking hours afterwards.
If you want to stand out with your service almost orthographic images are a great way, but I also suggest to pose your models properly.
Either both of your models have serious posture problems with their spine or you didn’t care enough.
Both lean a lot to the left and bend their head to the right.
Also the backshots are not entirely from the back.
So overall I find the quality lacking and I would not pay for it as I’d have a lot of work to do to refine the images for reference images.
I’d also like to have a backdrop ideally alpha channeled with a heightscale and horizontal and vertical gridlines for proper and easy alignment and scaling of the reference images from various sides.
@mocap
Thanks! I thought it’d be useful to have the arm out of the way when modeling side of the body.
@arexma
Thank you. A tough, but fair assessment. I hope the following explanation doesn’t come off as defensive, because I don’t intend it to be. When I set out to do this test photo shoot, my primary goal was to see if it was even possible to acquire an orthographic photograph of a person in the first place, since in my preliminary research, I hadn’t found any attempts to make one. I did some testing with Blender renders to test the theory, but eventually you just have to go out and hire (expensive!) models and shoot the real thing.
My secondary goal, of course, is to make sure the models pose correctly and consistently between shots, or, as you pointed out, it wouldn’t be worth the extra work to refine the images further in post. Among other things, I had a laser line projected up the center of the models’ bodies in order to get the spine straight and symmetrical. This process of alignment took quite a bit of time between shots, and, with limited time with the models, I had to do the best I could in the time I had while still getting the shots I needed to test my primary goal of orthographic images. Long story short, I’m developing better ways of achieving proper body alignment for my next shoot, because I do care about providing good reference material for artists such as yourself.
Arexma, you mentioned that you are a paying subscriber to 3d.sk. Would you mind if I asked why you like using them since (I assume) you are still a paying customer, despite the issues you mentioned? Are there any other frustrations you’ve had with using them or anyone else? In regards to my proposed service, would you consider orthographic images an important component to the reference images you would use when modeling? Or does it fall further down in your list of priorities? What would you consider the three most important factors for good reference?
Again, thank you for your honest critique, and I hope we can continue this dialogue so I can refine my product while it’s still in its developmental stage and provide the best reference material I can.
The laser line you use is sweet. Actually it would be amazing if reference images would be:
OpenEXR with layers
contain proper EXIF information
have the laser line on one layer
have some grid or height-reference like in mugshots for proper alignment
high quality closeups
one of the reasons why I was subscriber to 3d.sk - let’s face it there’s more reference there than one artist can use in a lifetime - is because it’s a huuuge library with lot’s of high quality images and close ups. I don’t know if you’ve seen one of their full image sets, there are closups of the inside of the ears, up the nose, genitals, the eyes looking up,down,left,right, the inside of the mouth, tongue from top and bottom, bottom of the feet, images of the hands with spread and closed fingers. Also sweet, the references where often shot with various outfits, underwear and completely nude (not all)
A paradise for projection textures or just texture making. The only real downside was that for ortograpic references you had to get your hands dirty. Usually there was a shot of the full body but quite perspectively. There usually were shots of the body as well in portions, head, chest, hips, upper legs, lower legs. You could stitch them together to a somewhat ortographic reference.
I usually worked after the perspective reference as good as possible and then switched to an perspective camera matching the FOV of the image (given the EXIF was there) and used the soft selection to match it. But only if I wanted the human to 100% represent the reference images.
Generally the requirement is anyways to make the human completely unreaslistic but “aesthetically pleasing”
So while ortographic-ish reference images are sweet, it’s not necessarily a requirement for every job.
Personally I know of no real competition to 3d.sk. So go for it.
