What improvements do photomodes make? *moved*

[Dravonic]

I'm new here and I'm not sure if this is the right place to ask but... well I hope it is.

I've been taking a look at GT5 "bullshots" lately, which appear to be from photomode, and I was wondering what kind of modifications are made to them in comparison to real time. Unfortunately I can't compare them to GT5 for obvious reasons, so I'm comparing them to GT5P.

My best guess is that they are the gameplay with added depth of field, motion blur and anti-aliasing. All the rest appear to be in-game like reflections, textures, lighting, models, etc.

I'm just making a wild guess so I'm searching for more info on the matter. Has this been discussed already? Have any conclusion been reached?

 

[jlippo]

From what I can see they do accumlation buffer AA, DoF and motion blur, very easy to do and you can archive incredible results.
Basically you render image over and over again and blend them additionally into a deep accumlation buffer.

Each time you
Jitter camera in a way that shifts location of pixels slightly to get AA.
Jitter location of camera within camera lense to get DoF. (camera target stays put.)
Jitter game-time to get motion blur.

More frames you render the better the image will look, in the end just divide the resulting buffer with the amount of frames you rendered to get the result.
Something like 120xAA takes 2 seconds with a 60fps game in a game resolution and the quality is way beyond any post processing effects in both DoF & Motion Blur. (or AA)

 

[Jedi2016]

My best guess is that they are the gameplay with added depth of field, motion blur and anti-aliasing. All the rest appear to be in-game like reflections, textures, lighting, models, etc.

As far as I know, that's exactly what it is.

 

[Arwin]

I'm not so sure that that's the only difference. I have a feeling (haven't looked at it close enough) that there's more to the lighting as well. If I'm correct that there is some form of ray-casting or advanced reflections going on using the SPEs, maybe photo-mode is doing that at a higher resolution as well - which would make sense if you realise that photo-mode in GT5 can render 8 megapixel images.

You already have gameplay, then replay mode, then the garage view (which seems to have more advanced effects than replay mode), and photo mode may well be a step up from that again.

 

[Cheezdoodles]

which would make sense if you realise that photo-mode in GT5 can render 8 megapixel images. .

I dunno about this, any photomode can theoretically render8 megapixel images. Offline rendering has allmost no limits on resolution

 

[Laa-Yosh]

If I'm correct that there is some form of ray-casting or advanced reflections going on using the SPEs, maybe photo-mode is doing that at a higher resolution as well - which would make sense if you realise that photo-mode in GT5 can render 8 megapixel images.

What would they do with raycasting? It's basically, fire a ray from the camera's view point along each pixel and see what it reaches first...

 

[sebbbi]

Each time you
Jitter camera in a way that shifts location of pixels slightly to get AA.
Jitter location of camera within camera lense to get DoF. (camera target stays put.)
Jitter game-time to get motion blur.

And jitter light positions to get soft shadows

Racing games are usually 60 fps. By using one second to generate the "photo" you can accumulate 60 frames. That alone improves the quality drastically. No more polygon edge aliasing, no more pixel shader aliasing, no more shadow acne, perfect DOF, motion blur and soft shadows. And the best thing is all you need to do is jitter various rendering properties. No extra code (or memory) needed to handle DOF, motion blur, antialiasing or soft shadows.