> I can envision tools that allow a sign in a store front to change from "Coke" to "Pepsi" automatically, or even replace that Audi car the hero is driving to a Mercedes. Product visualization/ads has had pipelines set up for this near drag and drop for static assets for a while now. Their assets are known before production, and have been decided by the agency that made them. Animation and Film VFX are setup fairly differently.
The ability to make changes like that already exist. Scenes and assets built procedurally can be altered with relative ease as long as they don’t encroach on other aspects of the shot.
> Specifically things like changing sponsor logos or animation models.
Logos and for the most part basic comp-only changes? Yeah that’s fine. Animation model is a completely different story. Changing the base model means the UV’s likely changed so the texture artists will have to either update (can be complex if they’re trying to maintain their original stack) or redo their work. It means the rig needs to be updated and modified. If it’s a muscle based system for precise and realistic skin deformation, have fun. If there character has fur, then the grooming team has to do another take on the asset as well. Animation then needs to make another pass on their scenes to make sure the animation actually looks good and lookdev needs to make tweaks to maintain the look they had before. And then you have to rerender, which will likely conflict with the rendering allocations (if using an on-prem farm) of the current deliverable unless the studio is in a lull point. And rendering takes time. Then comp gets to deal with the change and make sure it still sits in the shot. And we still haven’t discussed the time and tweaks to re-sim and cache any affected effects.
For a movie like Cats, this is going to be a lot of work, with a ton of anim, matchmove, and comp fixes.
> The original VFX team would just have to insert "markers" into the workflow which could then be replaced and re-rendered. Almost like changing build flags and recompiling a program.
That is a significant misrepresentation/and lack of understanding of how these pipelines work and function. No matter how much you can automate, this is not just data the way the typical HN user sees things. This is someone’s art and vision, and will always require a hands on approach with human eyes. If it was as simple as changing a few parameters or altering a few lines of code, there would be far fewer people in the industry.