Recruiting a programmer (paid) for a 3d project (CLOSED)

[Ryue]

Hi
I’m looking for another programmer to help me out with a small 3d part of my program. The job would be a paid one naturally as I know how much work 3D things can be. The exact price we can talk over after talking about the specifics.

A few infos about the project and me:
As a few may already know I’m a programmer myself and I want to include a 3d crawling engine like the old school rpgs had (eye of the beholder, most d&d games or even visual novels like raidy nowadays) into my program. I have created a few prototypes already, although the way I went there did not bring me the flexibility and easieness to use (especially in terms of artists drawing only 1 or 2 images for the tile graphics) I want.

So what is the target?
The target is to create a pseudo 3d engine (first person view) which can display quite a distance and places the tiles appropriately on the screen. For the artists it should be so that best 1, max 2 pictures need to be drawn per tile.
Why do I say pseudo 3d engine? Because I don’t need a real 3d engine. In the game I ONLY feature 90° (or 180° for fast turn arounds) turns and when one moves he moves 1 complete map cell in a go.
Another point to make it easier is that the 3d engine itself does not necessarily need to handle the turning,…. So it could be programmed in a way that allows only 1 direction to view (upwards, or right as example). As the map itself is a grid and I only have 90° turns I can easily copy the parts of the map that could be seen into a pseudomap array just for the 3d engine itself.
This also means that overlays like movement buttons are not necessary as I can do them from the mainprogram. The pseudo 3d engine would only have to do the rendering of the tiles on the screen, all other parts I can use from my current prototypes.

For more details or if you have questions feel free to say/ask here or via pm.

[Ryue]

As a little more info: Either experience or willingness to work oneself into with old school rpg crawler engines or directly raycasting would be superb (or if the programmer has another idea of how it could be done, I'm very open there as long as it can be done with python and thus renpy^^)

[Ryue]

As the question came up when I explained someone what I'm looking for there.
I want an engine that has similar capabilities to the one used for eye of the beholder in the 1980s. Even though renpy is slower than c an engine like eye of the beholder engine (a typical dungeon crawler engine) can function easily with renpy as the 1980s pc's were surely (even with a compiled program) not faster than todays machines using renpy.
What I want is though not the 3 distance steps engine like eye of the beholder but an engine that could do 5-7 steps instead (dynamical effects like snow, fog,.... I can add myself there as I know how to do them)


An eye of the beholder tutorial even exists if someone is interested in it: http://eob.wikispaces.com/eob.vmp

[LeonGuisti]

As someone with a background on rendering 3D images and experience in 3D game programming (mostly using Blender, 3dsMax, and Unity) I feel I have to warn you that perhaps renpy/pygame is not the best choice for your project.

Ray casting IS achievable on pygame which can run within renpy (or vice versa), but getting it to do so is a feat within itself—and trying to imagine importing those libraries into renpy... that just makes my head hurt. Renpy is incredibly easy and simple to pickup, but it's also very limited. It, like many other game engines out there, has its specific uses.

I strongly suggest you look into Unity, Maratis 3D, or even love2D which can handle ray casting a lot better than pygame, if you're truly that hard set on it.

And yes, you're totally right, renpy is far more sofisticated than what people in the 80s were using. But there's also software today that's infinitely better than renpy for handling projects like the one you've described.

[LeonGuisti]

Also...

Have you considered having a 3D artist map a grid dungeon and take screenshots of of each angle? This way, when the player moves, the corresponding image is loaded. It gives the player the semblance of 3D without all the headaches when coding it. Sure, you'll end up with 4 possible (5 if you cout looking up) images for each spot on the grid—thus about a million images—but if your UI is like Eye of the Beholder, the image is no more than 30% of your screen so the file sizes won't end up being too big.

[Ryue]

Problem what I see is that its not only 4-5 pics but quite a lot more. If one goes down that route (I tried that one already with self made pics btw) its a lot more as you would need to make 1 pic for each possible situation (1/2 of the possible situations to be exact as you only need left side until middle as the right side you can just "mirror"). The problem with that is that for 5 distance steps it would already be quite a lot of pictures: 20 pictures per tile and with that you burn out an artist quite fast (the spacec is not the problem but the artists burning out is). Although for filesizes don't forget that even though they are "relatively small" if you have a base 1366x768 then even the basewidth of a singlle image is 409 so if you have 20 images there can be quite a lot of memory usage (but like I said its not the main problem there).

Why do I say that I would need 1 image for each posssible location if I take that route is easy because if I don't I need to precalc or whatever the position again and then I have the problem I had quite a lot of times that the images are not fitting to each other (one is a bit too much to the right the next one to the left as they are not correctly scaled,...).

Or did you mean a different variant of this method there?

[Ryue]

As someone with a background on rendering 3D images and experience in 3D game programming (mostly using Blender, 3dsMax, and Unity) I feel I have to warn you that perhaps renpy/pygame is not the best choice for your project.

Ray casting IS achievable on pygame which can run within renpy (or vice versa), but getting it to do so is a feat within itself—and trying to imagine importing those libraries into renpy... that just makes my head hurt. Renpy is incredibly easy and simple to pickup, but it's also very limited. It, like many other game engines out there, has its specific uses.

I strongly suggest you look into Unity, Maratis 3D, or even love2D which can handle ray casting a lot better than pygame, if you're truly that hard set on it.

And yes, you're totally right, renpy is far more sofisticated than what people in the 80s were using. But there's also software today that's infinitely better than renpy for handling projects like the one you've described.

I just don't think that it really matters with what small scale I have in mind. In total it is about 90° turns and no up looking or down looking and "only" 35 different blocks that can be displayed if we use 5 distance steps or 63 blocks if we use 7 distance steps as maximum. So there should be something possible relatively easy even with renpy. Raytracing, pseudoraytracing or even something different. For example also one of the cookbooks has something going into this direction:
http://lemmasoft.renai.us/forums/viewto ... 51&t=19245
But it uses quite a lot of different pics for displaying 1 tile. And I want at max 2 better even only 1 pic only for a tile.

[LeonGuisti]

Problem what I see is that its not only 4-5 pics but quite a lot more. If one goes down that route (I tried that one already with self made pics btw) its a lot more as you would need to make 1 pic for each possible situation (1/2 of the possible situations to be exact as you only need left side until middle as the right side you can just "mirror"). The problem with that is that for 5 distance steps it would already be quite a lot of pictures: 20 pictures per tile and with that you burn out an artist quite fast (the spacec is not the problem but the artists burning out is).

The cool thing about 3D rendering is that once you've rendered out a model, you can take as many pictures of it, from as many angles as you want. Additionally, it's a million times easier to edit than ray casting because in ray casting there's too much math that can go wrong.

I did these two renders in blender. The first took me, no lie, about a minute. The second one only took me a moment longer because I couldn't find the proper textures (and get them to work). But all in all, I'd say about 15 mins. Now if I want to edit what direction the camera is looking towards, all I have to do is turn it in the model, and I have a different angle. It's the same model, but just in a different angle. If space isn't a problem, you really should look into this. You can also add other stuff on the floor (rocks, barrels, plants) that would take a while to do in ray casting with relative ease.

[Ryue]

Hmmmm I took a bit of time to mull over your idea there. I'm not sure still so I would like to clear it up a bit with you there.
The screenshot you made is "1" specific situation which you take pics at different locations and angles. It has a specific layout of the blocks and
floor and ceiling and horizon.

What I need to do is (as I have maps loaded from .csv files) when I go down this route: Have 1 block/tile that has 1 picture per possible location. The tile itself is shown but all else is transparent.

Then I check the map programmatically and include the player location and show the pics that are there. Then I myself add in the floor, ceiling, horizon (this is absolutely no problem as its in reality only 1 bg that is loaded).

Also this would need to be done per tile (thus for a canopy of trees tile, for an alleyway tile, for a door tile, ........)

So with a max. distance of 7 there would be 63 different (fixed) situations that must be recorded per tile (per tile, thus per graphic of the tile itself like canopy of trees,...).

Did you mean it that way or differently then?

On the note of memory usage,... I made a few calcs there for this path. For it to be no position calculations (which are the cause of most problems as we both know), the pictures would have to be the size of the VN resolution. In my case this would be 1366x768.
I took a look at a few different bg pics that I have. They are between 477kb and 1.21 mb in file size.
If I take this as memory usage and use for example 20 different tiles (in combination with the 63 pics per tile) I would have a memory and also file space consumption of: 586.93 MB to 1524.6 MB. And that would be only the pics for the tiles themselves (if I would make something bigger with 40 different tiles....ouch).

[LeonGuisti]

Yup, I definitely underestimated the amount of space it would be using. Ray casting would be a better choice here for sure. Unfortunately, my experience with ray casting is extremely limited. I tried to recreated the effect on love2D once after I saw a youtube video on how to do it—but I ended up giving up and created one in Blender and Unity instead. It looked a lot like the images I posted earlier.

I would suggest looking up some pre-made pygame libraries and hope those play well with Ren'Py. You could also ask another forum for help. If you're offering money, I'm sure there's tons of hobbyist that would jump at the idea.

Sorry I couldn't be better help.

[Ryue]

Its no problem. I know that raycasting is a bit complicated by itself (and good non program language specific tutorials are rare :/ ) I tried now for 2 months to come up with a good and flexible solution when I called it a day on that and decided I definitively need help there in order to make the engine as flexible as I need it (and as small in memory consumption).

Tnx for your input though gave me a bit to think about there and a few ideas. Tnx also for the links will take a look at them.

[Ryue]

So programed it out myself as of that this thread is closed