SyntaxBomb - Indie Coders

You forget why you would use the AI model generator at all: because you might not have a "clue" how your model should look, or you are not capable of doing the right "proportions" ... etc.

I guess if properly automated the process is done in < 1 Minute. Of course the initial step (trying out variations) will also take some time. Nonetheless you will not need the 12-24 hrs for a proper game model - not to talk about the generation of the uv texture at all.

Yes, for simple stuff you might be faster doing it on your own - but the more complex it gets, the more time you can save if there are specialized tools (and no, I do not think Fragmotion is specialized in all the tasks required here - simply do some shrinkwrap with original model and remeshed/lowpoly one, place UV seems as you like (if there isnt an AI tool for it too ), reproject the UV and done).

As said the stuff taking most of the time will be modelling, uv and rigging - and a lot of this can be "ai-based" - with the disadvantage of adjustments in the early stages (modelling, texturing) are less easy ("generate new/variation"-Button...).


bye
Ron

i don't agree from what i have seen so far...

the resulting mesh / texture generated by AI looks good, but it is not ready for video games, too many vertices and triangles, messy uvmap, bad mesh structure around joints, too many joints, etc...

but i am open to change my mind if i see better results.

As long as Wonder Woman or Lara Croft's mesh looks good around the joints and curves i wouldn't be to fussy or Hard on yourself.

Just Saying Like!

Im by no means an expert but im finding if something works for now go with it and get it working

You can perfect it latter

the trailer / summary videos generated by AI, that you showed us in another thread, are most likely precalculated and prerendered.

video games often need to have stuff calculated / rendered in real time.

you see the difference ?

it is fast to render 100 pumpkins which are low details meshes (few hundreds vertices and triangles), low details texture (few hundreds x few hundreds of texels)

but it is slow to render 100 pumkins which are (unecessarily) high details meshes (several thousands of vertices and triangles) and (unecessarily) high details textures (thousands x thousands of texels).

because there is also the bots AI, the collisions detection and repositionning, the physics, the pathfinding, the animations of joints/bones and skinned vertices, the others graphics effects, to calculate / render.

of course it is always possible to have a high end computer to calculate and render high details meshes and high details textures, but it is not a good idea because not everybody has a high end computer, and to be frank, as lone developers, we don't want to compete on graphics, it is a lost battle imo.
i prefer to create something with original unique graphics / ambiance, and original fun gameplay.

and for those who use Blitz3d, an old directx7 engine (incredible that it still works well in 2024 ! ), we can't render too high details meshes anyway, or there will be a MAV at renderworld().
end of debate 

Oh my i think i get you. In the early days i wanted to render 49152 pixels at say 50hz. Ive seen demos where they paste random tiles to make it look like they've done 72hz to look like the same thing?

But in reality i needed the 49152 pixels at 50hz

In the last few years we don't seem to be moving forward anymore? Were are the 512bit computers and the 100Tb drives

At one stage everything seemed to get exponential every year.

recently i have played some old school 3d video games which used 'impostors' ( flat rectangle + texture ) of prerendered 3d models with different poses / perspectives. ( doom, blood, shadow warrior, heretic ).
and to create such prerendered 3d models with different poses / perspectives, a.i. can be useful indeed, because it does not matter if the 3d models is high polys, has a bad uvmap, has a large texture, it will be prerendered anyway.