next »

[Armageddon]

Hey, so I'm facing a semi-large design decision and I'd like to get some outside input because one way will make it a lot easier for me but it might make the user experience suffer. So my game's main mechanic/toy is the ability to photography the world and my problem stems from if I want to tell the player what he's photographing or not.

I'll tell you my problem with making it visible first. My game there are objects that you can photograph, from a programming perspective these objects are tagged as being photographable and they have a string saved to them so NPC's can comment on them using that string as a key. One of the things that's scared me from the start is if that system ever fails your illusion and trust in the world being created is just shattered. Maybe you're taking the photo fro ma weird angle and it's obscuring the raycast to save the string, or maybe it raycasts to a different/closer object. Of course this can be fixed to a degree by just tinkering with the system throughout development. Anyways anyways, my idea to counteract this was to give objects a lowpoly wireframe when they're "in focus". Like this:



My problem with this is that since I'm currently only doing the photographing raycasting from the crosshair out into the scene, it maybe affect how people photograph stuff without realizing it. Showing them what's going to be captured will force them to center frame every object just to "collect" all the strings in the world so they can see everything every NPC has to say about everything. I'm afraid it won't encourage people to take pretty pictures. Again this could be fixed by tinkering, changing the way I raycast, maybe have it based on the distance the focal plane is, etc. But the player is still going to have to be in certain static positions of the world to photograph each object. Showing them exactly what's going to be captured though fixes my worries about the player showing an NPC a photo and having the NPC say the totally wrong thing.

On the flip side of this though, having nothing, well, as I explained it can fail, even with tons of tinkering. Is it worth spending so much time tweaking do you think? To have that extra bit of immersion. I mean the lowpoly wireframe compliment the art style I'm going for. But at the same time, since the main focus of the game is photography maybe I really should focus the time into it and make it feel as real as possible. Let the players do their own framing and stuff, makes it more fun.

I hope I explained that well, it's been difficult for me to articulate this problem. Any thoughts?

[Canned Turkey]

What I would do is give the player a menu to highlight the subject of the photo after they've taken it.
With the rule of thirds and creative framing, there will be more times where the subject isn't in the center of the photo.

[gimymblert]

What's your problem?

- capturing too much info on a photo? solution: salience (ex: angle x distance x size x closeness to framing points)
- feedback on what can elicit reaction? solution: subtle art cues

[Armageddon]

What I would do is give the player a menu to highlight the subject of the photo after they've taken it.
With the rule of thirds and creative framing, there will be more times where the subject isn't in the center of the photo.

That's a good idea but it'd be quite a technical feat from my skill level to map that stuff as areas/chunks on a 2D image.

What's your problem?

- capturing too much info on a photo? solution: salience (ex: angle x distance x size x closeness to framing points)
- feedback on what can elicit reaction? solution: subtle art cues

My problem is that if I do the wireframing the player will more than likely center frame all the photos and it will devalue the photography mechanic, but it'll be a helluva lot easier for me to program and it would never ruin the immersion.

The other problem I forgot to mention is wireframing big stuff like buildings and the sky/sun/whatever. That could look ugly so the wireframing isn't an end-all solution either.

[gimymblert]

By wireframin you are talking about a highlight mecanism? and that this mecanism would draws the player too much to optimize photo to interactable objects?


If so, then maybe the problem is a problem of "feedback deapth", then still use salience of ALL objects into the frame + a 2d analysis of the composition (for example object that are repeated, interesting angles, cropping of some part of some objects) + character personality who like different compositions (some object centered, cluttered, dynamic composition, use of third lines).

If all that is relevant maybe also look at this talk http://www.gdcvault.com/play/1015528/AI-driven-Dynamic-Dialog-through

http://www.valvesoftware.com/publications/2012/GDC2012_Ruskin_Elan_DynamicDialog.pdf

You'll get it starting p27

[Armageddon]

Wow that video is basically exactly what I'm doing. I guess what I really need to do is just focus on making a more robust/complex way of capturing strings from the scene. I have zero idea of how I could save the composition or tell if objects are cut off. Shooting wide vs shooting long is easy enough by saving the fov, I've thought about telling the brightness by using a shader to crush the contrast and then calculate the amount of pixels that are black vs white, that probably sounds dumb though.

But yeah in light of that video and it lays out the systems a lot better than I had, I think it's probably best to just work on the mechanic more than adding a lazy highlight effect.

[gimymblert]

Fortunately plenty of game have implemented photographic composition into their gameplay!

It's easy, basically save the camera setting of the frame + 2d object position projected on that frame (might be centered outside the frame though as the object center can be outside if cropped) + depth + orientation + eventually frame of animation.

- Using the size of the object, it's depth, angle and center you can compute the degree of overlap with the frame. Then using the description of the object you can try to find which part are on frame (given an object have tagged parts you can access their data). it's just shape overlapping.
- you compute 1 or many bounding shapes (or any other methods) and compute the distance of the center of  of the shapes with define line or point of composition (for example the third lines)
- animation frame is to reconstruct eventually where an animated part would be in the frame at the scoring moment.
- Many game use the brightness histogram to judge exposition, you can also maybe divide the image into many small square and compute a score per square, then look at the distribution of score into the final image per define region to guess an overall quality (contrasty, colorful, etc)...
- whatever float your boat lol, there is so many idea! However looking for occlusion might be a bit more difficult.

Those idea could also be computed at frame level and stored in the resulting image tags, I assume "minimal tagging" (that's debatable) on saving the image and scoring later on evaluation, might not be as smart as I think lol.