I’m trying to make reference images of a human, but they always seem off. I’m aware that they can’t be perfect, but I always seem to outline the front side of the head way too thin, yet my female model’s face doesn’t look as thin. I’m modelling her in orthographic and noticed that objects seem wider in that mode, than in perspective. Maybe that’s the root of my “thin” problem?
Here are some samples of some pitfalls I always fall into:
Notice how her right side is not outlining her face properly like the left. Despite having symmetry on, I guess it’s due to the angle I took the picture?
And on this one, the left side highlighted vert marks the center of her lips, but on the front it’s on the top of her lip. Again, I’m sure due to the camera angle.
Are there any tips you all would recommend in order to avoid these kind of pitfalls?
You can’t expect a close up photo to be close to an orthographic view, there’s a lot of perspective distortion so you make the face more narrow than it should be (if you’re modeling in orthographic view). If you model a narrow face (perspective view) in orthographic view it will end up doubly narrow when you then look at it in the perspective view in Blender.
Also the photos from both views seem to be different in size, if you draw horizontal lines from the bottom of the neck to the other side you can see that it lines up, but the mouth and nostrils and so on don’t line up so I would say the right side photo is too small in scale or you tilted the camera when taking the photo or something. Honestly the photo doesn’t look like it’s taken straight ahead, you can see the floor in the photo.
As for the symmetry thing you can’t expect humans to be completely symmetrical so if you’re modeling with a mirror modifier you can pick the side of the face you prefer and just try to follow that. You can try actually mirroring the photo in the center and see which way looks the best.
Anyway I would say just have the reference photos act as a reference for the proportions when making the face instead of trying to follow every detail when modeling. You can adjust it later when you’ve filled it in more to more closely resemble the model. But if you really want to use references like this you can take the photos a bit further away (if you’re modeling in ortho view), make sure it’s straight ahead by using a stand or something and make sure the distance from the camera to the model is the same. If you want to do my model-roughly-by-reference style you can try taking video reference instead, just film lots of different angles and you can make sure the mesh looks good (accurate) from all angles.
Wow very nice detailed answer, and yes you’re right, the pictures were not really taken straight ahead, or well…carefully. My biggest mistake was depending way too much on the reference image, which is something you mentioned in your 4th paragraph. I’ll try what you suggested and just use the reference images as a proportional reference and then fill in later.
Oh and mirroring the reference image is an excellent idea, I’ll give that a try too.
It’s not that difficult to get a set of workable reference heads for a person, and especially useful to paint from when texturing the head as well.
Use a camera with a zoom lens set to a fairly long focal length or a telephoto lens - the biggest problem is a wide angle lens, as in your pics above. The longer your focal length, the closer you are to an orthographic camera
Put as much distance between you and the model as possible that still lets you get a whole head and neck shot
Make sure your model looks straight ahead for every shot - the best way to do this is to use a stick or pole under her chin so the chin to ground distance does not change shot to shot, but it also works to have her look at the horizon - if you have one…
If you are going to use the pictures to texture the head:
Shoot on an overcast day to minimize shadows or in a studio with flat lighting. (If you don’t have either, use an on camera flash combined with the ambient light to fill in the shadows - but watch out for shiny spots on the skin created by the flash, you’ll have to remove those later)
Take 8 shots to minimize distortion, front, right, left, back and the halfway in between each of those
