-snip

This going to be good, i really expecting this kind of thread

Just my opinion but i feel like an analysis should go into a game once released, not based off a trailer for several different reasons

Pre-release analysis
I am basing this analysis off of the latest trailer here: https://www.youtube.com/watch?v=boBSQXVo5lU With it being a youtube video and not locally run software, while many of these points may turn out to be valid and present in the final game, keep in mind that youtube compression can and does play a hand in deminishing a level of quality, fluidity and color.

I will re-address all of these points once the game is released and in my hands.

Texture load issues
These are still present, and numerous even for the trailer, the most obvious of which is shown below, not all assets are affected which suggests to me that as in the below example, the console is loading all data for the scene that is static, then streaming in the dynamic portion of the scene afterwards, ability to change clothing, use a different character and/or set colors would be the more obvious reasons for this, as if the main actor in the scene showing this issue had a permanent art set then it could just be loaded in without issue the same time as the rest of the scene.

it's unfortunate that this is still the case after these issues being present on earlier videos, the extra few seconds or so required to fully stream in the texture could easilly be amended by adjusting the load order or clipping surface texture streaming, essentially its an issue, but one that COULD be fixed if they actually tried to, I would imagine that playing from disk or installing to internal memory would play a part in the overall load speed here, however it's feasible to assume that the trailer was made using a build running from internal memory rather than disk, so there is a possibility that people choosing to play from disk may experience this issue even more than is currently shown.

As you can plainly see, the difference between LOD low and high is considerable, with high having approximately 4x the detail.

Mesh replication = Nicer visuals
One thing that has passed over well from previous generations is the tendancy to duplicate mesh and texture data with instancing within the same overall model, doing so allows for objects that seemingly have a very high complexity but dont find themselves as resource constrained as models of similar complexity would be, because the instancing of materials and mesh data can be done with much less overhead and significantly less resource use.

For the wings of this machine, instancing of a single segment allow for replication across both sides of the model, creating a complex overall model, the replication is also seen on the head which is mirrored, the circular turbines and the bodywork itself, the barred fingers are then replicated four per side of the same mesh, overall the result is a very nice model with little more resource use than what is required to render out a few smaller parts, as the camera pans out (lower portion of the image) we see that this level of detail ends abruptly at the structure on which it sits, even though it has reflective replication in play, the overall structure is much lower poly and a stark constrast from the perceived complexity of the main machine.

We also see this repition of instancing throughout player models and actors, as illustrated above.

The most telling use of this is on the flyover shot of the city.

You can plainly see that in this LOD version of the city, traffic is an absolute minimum, tthere are only three types of tree, and many of the buildings are duplicates, the first row of houses next to the church for example is copied entirely on the far right row of houses, ll hedgerows and geometrical data of the city is removed, which contrasts greatly with the playable-lod version shown immediately after in the video.

Emissive surface lighting versus emissive realtime
One of the notable differences seen in recent videos is a stark difference in lighting, from my own experience with the studio the shift is from emissive textures that generate light that effects surrounding textures with reflectivity, to emissive textures that emulate it by effecting the lightmap only, the lightmap data is then treated in postprocess with a light haze that surrounds the emissive portion of the image regardless of surrounding geometry.

As we see in the above image, the bright pink neon looks nice enough, but does not emit light that is reflected off of other surfaces, instead the haze simply effects the entire screenspace evenly, ignoring mesh occlusion and depth order.

We can also use the arm in this image to confirm that the presentation is indeed 720P 30FPS, though some portions of the game are clearly struggling to maintain that framerate.

Rigging on cutscenes
Perhaps unsurprisingly, corners have been cut in cutscene animation, in the below scene that i've stabalized we see the gun moved independantly of the hand once lifted by the character, watching the video alone is enough to confirm this, the stbalized gif is merely to highlight it clearly.

The lateral motion of the weapon suggests poor bone rigging for the model, or thst the model was not even present when the characters animation was created, i am actually more inclined to assume that the weapons animation was added in after and played along side the characters animation rather than merged to it, the way the gun moves independantly of the character moves to confirm this theory.

Overall performance
At three points in the video, the framerate is greatly impacted, two of which are caused by explosions, one of which as a result of "too much on screen", specifically an ariel shot showing five dolls close to the camera and 17 dolls in the distance, the distant dolls are extremely low poly/texture versions, so if we are generous and assume that their combined complexity and texture data is equal to two dolls then we get a situation where showing 7 full lod dolls on a screen with only a flat textured surface (ocean) results in significant framerate impact, it is clear then, that the adjustments made to the earlier cutscenes where a number of dolls has been removed as well as simplification of the internal structural geometry, has been done because the engine could not handle these on screen without impacting framerates.

Geometrical differences on dolls
In the opening scene, the shot of the character inside a doll features relatively clean textures with pretty nice geometrical detailing, later on in the trailer we see an internal shot of the same doll, piloted by the same character, but with a much lower texture detail and reduced geometrical detail as shown below.

It is likely that two models for the same doll exist, or that in particular cutscenes the model is used with a lower lod level forced when the scene will potentially be more complex, as shown on the left you see the see passing by and fellow dolls flying in, where as on the right its just a basic background, all of this further solidifies the idea that the engine is struggling with the job of rendering the game, and cost cutting measures are needed to maintain a decent framerate.

Particle and blur effects
A welcome plus side from the corner cutting measures and strict mesh/texture diet is that Monolith has been able to employ a per object motion blur in cutscenes which do an excellent job of hiding the rougher edges and gide doll battles a nice feeling of fluidity and speed, something a lack of blur would lose, the particle effects employed are nice too, with volume based smoke used in conjunction with the emissive light postprocess to great effect, the motion blur for larger scenes also uses a somewhat bizzare approach of not blurring every frame, leaving some frames without any blur at all and others with heavy blur, these blur difference and smoke effects can be seen below.

Conclusion
While I will go back and address all of these with the final release, it's clear to me that the inexperience with working with more powerful hardware has had a significant impact on Monolith, a lot of corners have had to be cut in order to make the game work with the engine they have, I hope that in time Monolith will get to grips with the hardware better and improve the performance of the engine to better utilize the WiiU's hardware, while the WiiU isn't exactly a powerhouse, the performance of this title is below par by the hardwares own standards.

I'm not one to really mind visuals in a game too, but I did want to give this a read to learn a thing or 2 about the game and its development. You did confirm what I thought looked weaker in the game, which was the view of the town. It was good to see the small oddities in the video, such as the neon pink-ish glow on that machine on the girl. The lack of glow did bother me once I saw it. Over, it was an interesting read and this thread was appreciated. Good work.

I still stand by what i said.  Just wait until you get the disc version.  I already dislike the overly nitpickiness that plagues forums today, but when you point out the bad on a product that may or may not be finished, and compressed for youtube or other video sites it becomes even less representative of what may or may not be the final product.  To me i find that unfair for the devs who have worked on the game

I congratulate you on the work you did for this thread, but id still wait until you are able to play it on your own screen and when you know it is finalized.  Its a good idea though, but i still think it should wait until you are playing them yourself

Very good thread though

So are u saying that, Wii U game can still be improve in the future (exm: Zelda U)?

Can we request more thech analysis on other games like Killzone Shadowfall or Ryse, Bloodborne or games that have not been discuss before on digital foundry , or maybe Final Fantasy 15

That's what I was thinking. I was under the impression that X was pushing the Wii U to its limits. Sounds kinda weird that there is more for the Wii U to imput than this.

I'm not really interested in the game but I always enjoy reading you posts, threads and analyses OP.

I'm still reading and I wonder if your bluntness will be misinterpreted by people who only expect sugary sweet things.

No, the limit is always on developer, hardware maybe limited but great developer can make the limited hardware more precious or make powerfull hardware useless , and for graphic Santa Monica and Naughty Dog is the best.