Weapon Modelling Questions

[lowlyguns]

Hi, i have some question regarding weapon modelling in blender.

So i have created a high poly assault rifle that has 1.5 million verts, i am just wondering what is the normal way game creaters get this down for use in games.

If i use the smooth tool then apply the subdivision modifier i can get it down to around 400k verts but im not sure if that is really the best way.

Another way that i have just learned about is using the low poly version of the gun i have and then baking the high poly normal map onto that. Is this the most used way of doing it and if so do i just need to bake the normal map and none of the others?

Thanks

[Zenogias]

The second choice is what we usually use it.

Yeah ,you have to retopologize them and bake the normal map.

[The Renderer]

Yes, the usual way would be to make a low poly version and then to bake a normal map from the high poly model. Never did a model for a game myself but depending on the game engine you would want to go <10k polygons for a weapon that's detailed and in the player's face (like for an FPS) and even lower for a third person game.

Not sure what you mean by "others" but which kind of maps you need depends on your game engine. Of course your other maps have to fit the low poly model since that's what will be used by the game.
(Edit: If by others you mean the ones you can chose in the bake menu options, then yes, only normals)

[Salt_Extractor]

The first person view always uses high poly models, with normal maps for even batter look.
Its the first thing a person sees when some1 looks at a FPS so its really important for it to look pretty. In 3rd person games its not as relevant.
LoDs ofcourse need to be applied respectivly to all the gun models, a lot of games often have only 2 or 3 LoDs for guns (First person and 3rd person or First person and 2 lvls of 3rd person LoD)

The end model shuld not ever have any mods on it apart from the armature.(all assuming you're useing an engine differante than blender)
Prefrence is also for the model to be only 1 object if possible to avoid bugs and such.

Normal maps are generaly added to everything, programmers (for whatever reason) never give options to disable them; so the preformance change must not be ever noticable.

[Mr Chappy]

Originally posted by ihavenowingss:

The first person view always uses high poly models, with normal maps for even batter look.

They use higher poly models, not high poly models! There is a big difference, high poly models can range from 100,000's of tris to millions depending on the model...
In modelling terms nothing in a game engine is high poly, if it was then the game would not run. High poly in game terms was around 25,000-ish tris for a full character model, though this increases over time(in sync with better average user hardware) and may be more than this now, though still below 50,000 at a realistic guess.

Originally posted by ihavenowingss:

Prefrence is also for the model to be only 1 object if possible to avoid bugs and such.

This is a very broad statement and as such should be disregarded.
How many parts your model should be comprised of is entirely dependent upon how it needs to function within the game!

Originally posted by ihavenowingss:

Normal maps are generaly added to everything, programmers (for whatever reason) never give options to disable them; so the preformance change must not be ever noticable.

Normal maps are a texture. Textures in games have a setting, high/medium/low etc. Normal maps are included in this.
Turning textures off altogether would result in a really awful looking game, this is why there is no option to turn them off. Instead you are given the option to lower their size(quality) and therefore the amount of RAM they use, which improves performance.

@lowlyguns do what @Zenogias said and learn how to bake a normal map from a your high poly model onto a low poly version of your model. It's not an easy process and takes time to learn how to get right, you can get lucky... You can also get unlucky.... It all depends on your model.
Set your entire low poly mesh to smooth shading before baking, you will save yourself a ton of time and misery by using this method rather than smoothing groups with hard edges!!!

[lowlyguns]

Thanks for your answers guys, it has helped me alot in understanding this.

Another question, is retopologize just making the low poly from the high poly? And is it better to do that rather then making the low poly first?

Thanks

[Zenogias]

It depends on your workflow.

If I am going to create some of simple table top,I just create them with a simple cube and rescale them,after that I would create a dupplicate of that model and create bevel on the edges.
Then bake those normal map to the original.

So,you don't have to retopologize that model the whole.(It's pretty pointless to retopologize a simple model like that but it can.)

but If you are going to create a character you did better do your best with sculpting and create a new one with lower detail with retopologize method,then bake the normal map as the same.

[lowlyguns]

Thanks,
Another question is when you do bake how are you meant to find the right ray dis? When i do bake my weapon i always get that mustard yellow stuff on the normal map and it is really frustrating. Am i meant to go less or more and does it matter if that yellow stuff is on the normal when i apply it?

[still__alive]

Depends where it is (if it is on a part of the texure with no uvs who cares), but a normal map is generally supposed to be that blue color, like these: https://www.google.ca/search?q=normal+map&client=firefox-b-ab&dcr=0&source=lnms&tbm=isch&sa=X&ved=0ahUKEwj3hIiWpPzXAhVKymMKHYKJDbkQ_AUICigB&biw=1198&bih=742

If yours doesn't look fairly similar to those in terms of color, there might be something wrong with it. Here are some more links related to normal maps that you might find useful:

https://www.youtube.com/watch?v=0r-cGjVKvGw

http://polycount.com/discussion/81154/understanding-averaged-normals-and-ray-projection-who-put-waviness-in-my-normal-map

http://polycount.com/discussion/107196/youre-making-me-hard-making-sense-of-hard-edges-uvs-normal-maps-and-vertex-counts

[Salt_Extractor]

Originally posted by Mr Chappy:

Originally posted by ihavenowingss:

Prefrence is also for the model to be only 1 object if possible to avoid bugs and such.

This is a very broad statement and as such should be disregarded.
How many parts your model should be comprised of is entirely dependent upon how it needs to function within the game!

A weapon can pretty much always work as 1 object, making it so; cuts down on some of the freedom you have (in certan engines) when doing materials, but it avoids any derpy bugs that may occur because of colision or whatnot.

Originally posted by Mr Chappy:

High poly in game terms was around 25,000-ish tris for a full character model, though this increases over time(in sync with better average user hardware) and may be more than this now, though still below 50,000 at a realistic guess.

3 years ago we would cinsistantly try to keep it 30k trist around the player, in reality it was rarely even close to that number tough.

Originally posted by Mr Chappy:

Normal maps are a texture. Textures in games have a setting, high/medium/low etc. Normal maps are included in this.

Occluision maps are a texture aswell, but you will find options to disable them. Anything that works with light can be treated as preformance intensive. I see no reason not to be able to disable just 1 slot. But theres never options to disable this one.

[still__alive]

About disabling the normal maps - I don't think you have to worry about that, if you don't want to use one, then probably just don't include one. I don't think you'd have an issue if you just ignored including a normal map.

Or I suppose you could probably include a "flat" or blank normal map that doesn't change any of the lighting/detail information of your mesh. If for some reason you had to include a normal map but didn't want to.

[Mr Chappy]

Originally posted by ihavenowingss:

Originally posted by Mr Chappy:

This is a very broad statement and as such should be disregarded.
How many parts your model should be comprised of is entirely dependent upon how it needs to function within the game!

A weapon can pretty much always work as 1 object, making it so; cuts down on some of the freedom you have (in certan engines) when doing materials, but it avoids any derpy bugs that may occur because of colision or whatnot.

Collisions are completely independent from the model, you never use mesh colliders for weapons. You use primitive colliders or a collection of primitives(compound colliders). Mesh colliders are far too heavy performance wise for non static objects, and even most static objects(like trees).

Originally posted by ihavenowingss:

Originally posted by Mr Chappy:

Normal maps are a texture. Textures in games have a setting, high/medium/low etc. Normal maps are included in this.

Occluision maps are a texture aswell, but you will find options to disable them. Anything that works with light can be treated as preformance intensive. I see no reason not to be able to disable just 1 slot. But theres never options to disable this one.

Occlusion maps are totally different. Normal maps contain the baked data from the high poly model, without this you have a awful looking low poly object, they are essential for any high fidelity game.
Occlusion maps are for slightly better shadowing, something that can easily be switched off without any major issues.
They can also be merged into the albedo map to save on RAM and therefore performance whilst getting pretty much the same effect, or just added as a post processing effect cutting out the need for these maps altogether....

[Salt_Extractor]

Originally posted by Mr Chappy:

Originally posted by ihavenowingss:

A weapon can pretty much always work as 1 object, making it so; cuts down on some of the freedom you have (in certan engines) when doing materials, but it avoids any derpy bugs that may occur because of colision or whatnot.

Collisions are completely independent from the model, you never use mesh colliders for weapons. You use primitive colliders or a collection of primitives(compound colliders). Mesh colliders are far too heavy performance wise for non static objects, and even most static objects(like trees).

I know nothing about this stuff I'm just posting what I get as instructions most commonly.