@smhrbst@rickksntss They mentioned this at a q and a - something about the look of un cropped raw negative. The sharp edge is typically a mask? They wanted these to feel like you were seeing all that was captured. And was it every era or just the past?
@timsoret was it you i remember going through how to properly convert color to grey scale/ luminance and retain accurate vibrancy match? I remember finding it fascinating and can’t find the post
Great tip. We do this a lot when working in animation ( VFX lighting/comp ) as well for similar purposes; to check luminance values, making sure the darkest/brightest areas of a frame are in balance.
@HammerFist720@duckywriting Not images, no prompt involved. It’s a system build specifically to aid an artist. Not what we now consider gen ai, though i understand it can appear that way.
@realvodrin@hxvenangel@Wow35551 Katana is just Imageworks in house lighting look and scene build software. It’s not ai in the modern sense. It’s been around for 20 or more years. Though maybe now they e incorporated ai into it somehow. Katana was revolutionary back when i worked there in 2006
@realvodrin@hxvenangel@Wow35551 I do. Imageworks made this tool before LLM tech existed in a useable way. It was 2016. It is not a “model” in the sense that we understand today. I’m sure their tools have evolved now. But what we’re seeing here is a decade old.
@bitcloud Honest question. Is it a Lora if it’s not using a base model and outputs 3d line data and not imagery? My understanding of l that that means using a base model and those didn’t exist publicly when spiderverse was made .
@defnotbeka@chhopsky It would be but it’s not possible. It requires too much data. Billions of images. That or its based on a model that is already trained on stolen data.
@defnotbeka@chhopsky Sadly gen ai has become too general of a term to mean anything. The spiderverse way IS ethical in a way that diffusion transformer teqniquescare not.
@GabrielValles I guess i don’t understand what the stakes are from your point of view. I only know that diffusion models stole the work of friends and colleagues. Sell that work back to them in the form of tokens and call it democratization. There is no such thing as ethical image gen.
@aragon1500 People use “gen ai” to speak about diffusion models. Transformer tech. Spiderverse did not use diffusion models for this purpose. The tech wasn’t close to ready when this, the first film was being made. It isn’t the same tech.
@LobbyLoitering@fableshowrunner@Gossip_Goblin Came here to say this. There are rules. Animation, hand keyed animation is required in the majority of the film to be considered. Even novel films don’t qualify