Will we have a [decent] end to end AI generated anime series by the end of 2029?

@firstuserhere made an interesting market series here but the resolution criteria are very forgiving. This market attempts to answer the same question with less degrees of freedom.

Market Resolves YES Requirements:
- All relevant aspects are AI generated: Script, animation, music, art, voices, sounds, etc.
- One coherent storyline that goes across 3+ episodes.
- Episodes are at least 20 minutes in length.
- Quality: I personally consider the quality decent or reasonable.
- Characters express movement and animation consistent with anime that would be aired on Crunchyroll. (not neccessarily big budget)
- Primarily uses 2D animation style. 3D models do not count.
- Should appear to have a standard "anime" style.

If the above are not met this market resolves NO.

I will NOT bet in this market.
I plan to resolve based on the spirit of the question.

(I will slowly be making a series of these with different timelines)

Get Ṁ1,000 play money
Sort by:
predicts NO

@SneakySly can you give some examples of existing anime that are on the edge of your quality standards for this question?

@LukeHanks People can throw potential corner cases at me and I can explain my thought process based on those.

An example of not counting comes from FUH's initial markets here: https://manifold.markets/firstuserhere/will-we-have-end-to-end-ai-generate-5927e7f2d735?r=U25lYWt5U2x5#xDJqn9uHbHJp4V3Jb4Oy

His example fails on quality, and not being an anime stylistically.

predicts NO

@SneakySly I don't know what anime you've watched, and I don't want to burden you with a bunch of viewing assignments. Do you have a profile on AniList or MAL?

I guess I can take a shot in the dark. Is there a standard for narrative quality? For example would it matter that Domestic Girlfriend and My Life as Inukai-san’s Dog have terrible stories?

@LukeHanks Not up to date and probably inaccurate in many way, but I do have a MAL I don't use much:

I have not watched either of those examples unfortunately.

predicts NO

@SneakySly Thanks. You have some good taste. Would the lowest rated items on your list pass the quantity test if they were made by AI?

@LukeHanks Yes. The only one that potentially wouldn't would be the Spider anime as they heavily used 3d models. These all had coherent (if bad stories) and an animation level that clears my bar.

The definition of end to end AI generated is tricky, all the elements you listed, can be manipulated by human hands to increase the odds of good output, then output repeatedly until good material is generated, then assembled by hand separately and still all be AI generated. How does prompting work, one prompt for the whole series script, for one episode, one scene, for the entire thing all sound art animation etc?

@Paul7bbc Based on FUH's original markets, this does not need to be a single model but could be prompts to multiple different models and then stitched together. For example, you could prompt to a script generator and then submit that outputted script to a model that will create the visuals, etc.

@SneakySly every high quality AI video project I've seen and a lot of the better AI art, have used AI art every step, but the amount of human labour, LORA, inpainting, editing, selection, etc has been like rolling dice and only selecting sixes for hours and hours to reach the desired quality. I still bet yes in any case, because by 2029 "decent" would still be possible by whatever counts for autonomous AI, with or without substantial human curation.

Related market:

More related questions