Yet it sported better details like shadows and such (discussed here) not present in ingame play. Hence yes realtime but not fluid realtime and then recorded... hence pre-rendered.
"Yes, but how many polygons?" An artist blog entry with interesting numbers
- Thread starter Farid
- Start date
Yet it sported better details like shadows and such (discussed here) not present in ingame play. Hence yes realtime but not fluid realtime and then recorded... hence pre-rendered.
That wouldn't necessarily help. Most of our character models are between 30 and 150 thousand quad polygons before tesselation, so if a game were to use similar approaches then they'd still require lower resolution discrete LOD models.
Tesselation is not some black magic stuff... it won't add detail in itself, and even if it's combined with displacement mapping, high res characters will still require a certain amount of polygons pre tesselation for animation (be it bones and skinning or blend shapes or anything else).
Only if the game wasn't limited in anything else, like fragment processing, AI, physics and such... vertex processing isn't always the bottleneck.
Betanumerical
Veteran
What is the point them discussing polycounts, in relation to how they are using the ps3 well if they arent even real time numbers? theres no reason for them to lie.
COPS N RAPPERS
Regular
it isn't about adding more detail to polygon models, it's about subtracting them better (when needed.) starting with high LOD models and letting them fluctuate would aid in better memory usage, and that in it's self would aid in better performance.
AI can be expressed in different ways and can only be complicated if there's more you're working with (that's for both situational and graphical), same thing with physics, so in essence graphics in a way will always be a bottleneck.
A developer has the choice to give you any kind of figure for anything and it doesn't have to be exact, like turn10, or Capcom, capcom advertised Resident Evil 4 a lot and boasted about it's cutsceens being so smooth and interactive at the same time, but they weren't realtime...not a single cutscene in RE4 was, and that goes double for this new batman game.
some developers are like magicians, and do all magicians like giving accurate figures of what they do for a living behind the scenes........no.:smile: and you can't knock them down for that.:smile:
Last edited by a moderator:
Ether_Snake
Newcomer
The Prince of Persia 2008 main character had around 40,000 triangles And this was at all times, cutscene or not.
And this was at all times, cutscene or not.Except that the Batman game is infact totally pre-rendered while the Uncharted ones are pre-recorded so as to hide loading times. Also Resident Evil 4 Gamebuce version ran every cutscene in Real Time.
As everyone knows a good few games add additional details in cutscenes while running them off real time, ND said that they only used higher quality lighting & shadowing for their cutscenes & kept rest assets like poly count & object details the same. And from what I have seen they have been quite honest about what they say everytime. IIRC I remember there was a cutscene model vs gameplay model comparison somewhere & there was no difference in them.
COPS N RAPPERS
Regular
The Prince of Persia 2008 main character had around 40,000 triangles
I somehow doubt that.........you're talking about this game right?
http://i35.tinypic.com/2uszar5.jpg
http://i35.tinypic.com/2rhw5kh.jpg
....i don't think they were 40k.
Yep, cutscene models and the realtime models are all the same explained from the making of uncharted videos. So in Uncharted 2's case 80k main character model is something monstrous in a third person action. Still, Lair tops it by having 150k ridable dragon and the overall poly count on screen.
In Cel Shaded games most of the shading comes from actual models rather than normal maps & such, hence they generally tend to have a lot of triangles as compared to other non cel shaded games.
COPS N RAPPERS
Regular
...........you guys are doing a good job playing with my head.
....that sets it above mass effect1 or uncharted drakes fortune.
...but i could be wrong, lets ask (Laa-Yosh) since he works with graphics too.
Oh wise one, Is Prince of Persia's character models more than uncharted?
we anticipate for your blessings.
as for me i will pass on some interesting info. RE5's characters are about 16k and Street fighter 4's are about 16k too.
Google translate it.
http://www.softimage.jp/user_case/biohazard5/index.html
http://www.softimage.jp/user_case/street_fighter4/index.html
I'm sorry, but... what??
Tesselation is nothing else but adding more polygons. Taking away geometry is a different thing that isn't really worth doing in realtime, it's better to manually create discreet LOD models.
Sorry again, but what are you talking about??
And the rest of your post is not even remotely comprehensible...
Lair has continuous realtime LOD and most of the times you only see a fraction of that 150k model.
As for Uncharted 2 and 80 thousand polygons per main character, I find that very very hard to believe.
It wouldn't surprise me if they are counting polys for final result after being re-rendered to achieve desired effects. The Lair numbers would be believable for the cutscenes.
Both are pre-rendered with respective engine, both has better IQ, both hide loading times.
The Prince of Persia 2008 main character had around 40,000 triangles
Re-rendered several times to achieve visual results or base mesh?
I know their adaptive mesh system handles some strong LOD and you still do get to see that 150k poly dragon whenever you engage 1 on 1 battles which I do a lot. Thing is you can say that for most of the games when handling LOD, I'm sure they use a much lower res model when not zoomed in. Still that's a pretty big achievement in Lair for crunching out that number though.
I'm not sure what you mean by re-rendered, if going by your method they could've render a 1million poly drake no?
Really? And how exactly can you be sure of that, did you count them or what?
Their overly ambitious tech goals were one of the main reasons that lead to the game's failures. It actually made the game look a lot worse then other titles sticking with more reasonable tech.
That a object l has to be rendered twice or more due to effects applied to them. In Crysis for example base mesh polygon count is around 15-30k but ingame with effects applied they are 2-3x that amount.
So the question is 80k for base mesh or for final output, pre-rendered cutscene and/or ingame?
As reference these models are around 40-50k polys. Cloth curves are polygons and not mapping and it got separate clothes on top body mesh etc.
This and this
Last edited by a moderator:
If that part is the most up close view of the dragon with the highest LOD and according to Dev's own words then one can assume so. Also the cutscene model bears exact same look from what I can tell so that leaves to the CGI model.
Admittedly their engine lacks consistency in some scenes while some levels were totally spectacular, as well as considering the nature of any PS3 prototype engine at the time, I for one would thirst for an updated and refined lair engine.
http://translate.google.it/translat...ted-2-a-milano.php&sl=it&tl=en&hl=it&ie=UTF-8
So only a filter was applied, I mean I'm having a hard time believing it myself too so you tell me dude.
The crysis rendering method is confusing to me though, what kind of effect would require such a drastic increase in base model poly count? Wouldn't that cause major visual discrepancy when played?