88
Following accusations, James Pond Legacy devs say no generative AI was used in creating artwork (gonintendo.com)
submitted 1 day ago by to c/games@lemmy.world
50 comments fedilink hide all child comments
you are viewing a single comment's thread
view the rest of the comments
[–] 99 points 1 day ago* (last edited 23 hours ago) (8 children)

I feel like they did a shit job tryinf to defend themselves if this is the best example of steps taken:

RxvUGBFyGR7HDnN.png

Like, For Real, this is worthless. Show the messy work files, the unfinished sketches, isolate the layers in the 2D mockup to show lineart and color independent of each other, a gif of the 3D wireframes rotating.

This image they posted won't convince anybody. It will vindicate critics.

  • permalink
  • fedilink
  • source
  • hideshow 16 child comments
    •
  • [–] 67 points 1 day ago* (2 children)

    Yeah, I really don't want to falsely accuse anyone but... the human 3d-artist forgot to model one of the horns? That redraw has a very, let's say, familiar style as well.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] 13 points 1 day ago (1 child)

    Says “process” in the title unlike the others, body is missing inside, flame jets are missing as well.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 12 points 20 hours ago

    Body, flame jets and electric textures can be explained as they're not part of the 3d model but effects applied to it in-engine.

    The missing horn though screams AI and I have a hard time imagining a human artist forgetting one of the horns when making the 3d model and submitting the work as completed

  • permalink
  • fedilink
  • source
  • parent
  • [–] 41 points 1 day ago (1 child)

    i... these are literally all ai. like why would the render be missing a horn?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 11 points 1 day ago (2 children)

    I'm gonna interject and say you generally model things one half at a time and then mirror for symmetrical objects.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] 38 points 1 day ago* (1 child)

    Sure, but you wouldn't just render one horn and add the other one in photoshop. That makes no sense at all. You would duplicate the horn before rendering it.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 4 points 15 hours ago*

    I'll play devil's advocate and say there are non-AI render processes that can miss a horn (e.g. working from a clay model and scanned, for instance). Hell, even a lazy wand picker in Photoshop can lose a horn! I'll even say that the very-AI looking 2D illustration can simply be a byproduct of a lack of creativity and the fact this is clearly shovelware. I lived through the Wii era, I can attest that shovelware can look this bad without AI, lol.

    HOWEVER, it sure as hell looks like AI generated artwork and they could do a lot better defending their process. The irony is ... I don't really think the trailer is AI? It's... honestly too unpolished and weird to be? Every effect used is something I'm certain I could do myself, lol.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 32 points 1 day ago* (2 children)

    The theory on The Other Site is what I also agree with: They took #1, fed it into an AI 3D modeler to get #4, fed that image into a 2D style AI model to get #2, and then went back to #4 and used AI in-painting and lazy Photoshop to remove some elements to get #3.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] 11 points 23 hours ago (1 child)

    oh yeah that makes sense, the little bubble window is literally just painted over.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 5 points 19 hours ago

    As well as the weird dark blob where a detail is added in the "final" result. Why would the blob be there? Unless they started with the "final" result and painted it over to get to the "previous" step.

    That and the mangled udders clearly scream AI to me. The original clearly has 4 rockets intended to look like udders. But because of the perspective two of them seem to overlap. So the AI just combined them into one, with a sort of double output? This makes no sense, that's not how rockets work and that's not how udders look. It is exactly how AI works tho.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 19 points 1 day ago

    How is this not AI?! The redraw absolutely smacks of that style, not to mention the cannon arm is obviously meant to imitate a milk can which the redraw mangles. The rockets are supposed to be udders but the two we can see most clearly don't match in construction and the other two seems to sprout off of another one. These errors carry over to the full render. The udders are a pixelated mangle of perspective in the original, but this point is absolutely missed.

    Above the leg is also a weird split. You can see in the original this is a weird artifact of pixel shading, but it's been rendered into an actual split in the material in the redraw and renders. This is a valid design choice I guess, but to me it's an indication of how the AI misinterpreted intent and someone reviewing it didn't catch or care about it.

  • permalink
  • fedilink
  • source
  • parent