Sub-Project CSMs - Character Select Models on the CSS

[DRGN]

This is something that I've thought about for a long time, yet haven't had the time to investigate myself. It's the idea of using character models instead of CSPs.

The benefits I see include:

• Dynamically see each costume that you have in the game, without needing to create new CSPs after importing new costumes
• Free up a TON of space in MnSlChr (the character select screen file); and by a ton I mean over 3 MB, or ~80% of the file!
• Use the extra space above for more type _6 textures for the closed ports (which are pretty large), so that each one can be unique
• Have a really cool CSS

I think this should be mostly if not entirely feasible. The developers already started work on this as a feature, to be just like how they are in Smash 64. (They likely didn't finish it just due to time constraints.) We've already seen these turned on from Itaru 's work, here. However, what I'm proposing is to disable the animations (or make them 1-frame), and just have the model appear without moving. Of course, we would need to make some new poses too, which can be done with Smash Forge.

Challenges I see would be:

• Somewhat optional: To make a "mask", so that parts of the model that go outside of the CSP area are not rendered. Not only would this allow for close-ups, but more importantly, when you have a pose for a character that works really well for a portrait, sometimes it has a part of the model that's contorted and doesn't look very good. With a standard texture for the portrait we would of course just crop that off and use the good looking part for the portrait. So doing that with a mask removes the limit on needing to show their whole body and I think on average will give us much better looking selections. Not sure how difficult that would be to do. Maybe something like a formula injected into the rasterizer at the point of hidden surface removal that only applies to these models? Or create an object in front of them (with rectangles cut out) that is invisible, but which the rasterizer thinks is opaque when comparing against thees models.
• Figure out why some of the surfaces of the models don't render correctly and fix them.

Thoughts?

For step one of this, Itaru , how did you enable the models and the "Selected" animations to appear on the CSS?

 

[JoshuaMK]

I fully support this

 

[Itaru]

This is something that I've thought about for a long time, yet haven't had the time to investigate myself. It's the idea of using character models instead of CSPs.

The benefits I see include:

• Dynamically see each costume that you have in the game, without needing to create new CSPs after importing new costumes
• Free up a TON of space in MnSlChr (the character select screen file); and by a ton I mean over 3 MB, or ~80% of the file!
• Use the extra space above for more type _6 textures for the closed ports (which are pretty large), so that each one can be unique
• Have a really cool CSS

I think this should be mostly if not entirely feasible. The developers already started work on this as a feature, to be just like how they are in Smash 64. (They likely didn't finish it just due to time constraints.) We've already seen these turned on from Itaru 's work, here. However, what I'm proposing is to disable the animations (or make them 1-frame), and just have the model appear without moving. Of course, we would need to make some new poses too, which can be done with Smash Forge.

Challenges I see would be:

• Somewhat optional: To make a "mask", so that parts of the model that go outside of the CSP area are not rendered. Not only would this allow for close-ups, but more importantly, when you have a pose for a character that works really well for a portrait, sometimes it has a part of the model that's contorted and doesn't look very good. With a standard texture for the portrait we would of course just crop that off and use the good looking part for the portrait. So doing that with a mask removes the limit on needing to show their whole body and I think on average will give us much better looking selections. Not sure how difficult that would be to do. Maybe something like a formula injected into the rasterizer at the point of hidden surface removal that only applies to these models? Or create an object in front of them (with rectangles cut out) that is invisible, but which the rasterizer thinks is opaque when comparing against thees models.
• Figure out why some of the surfaces of the models don't render correctly and fix them.

Thoughts?

For step one of this, Itaru , how did you enable the models and the "Selected" animations to appear on the CSS?

That is Video editing lol
I just replaced Wait1 and Wait2 with Selected and SelectedWait, and recorded with black back ground.

 

[UnclePunch]

That is Video editing lol
I just replaced Wait1 and Wait2 with Selected and SelectedWait, and recorded with black back ground.

haha that tricked so many people

DRGN
I've briefly worked with loading and displaying models,

so assuming the preloaded model and animation data can be utilized on the CSS, the biggest problem i see is

1) the delay between hovering over the CSS icon and displaying the model (currently the character data is only loaded when selecting the character)
2) the delay between changing costumes and displaying them

i kind of agree with the developers that instantly seeing the character + costume makes for a smoother experience but its a very cool idea to pursue for the hell of it.