New animation! (not the one your thinking - 1m10s)

Caution! ⚠️ This animation is AI assisted but I am not moving away from blender. This is purely an experiment, mostly to entertain my own curiosity. It's a compeltely off shoot project.

This content isn't available on Fanbox.

Go to the Welcome post for more information.

No I haven't gone senile and uploaded the same animation twice. This is the result of some experimentation. To cut to the chase, this animation is AI assisted. Before that sets off any alarms, I'd encourage you to watch the animation first, then leave your comments.

This was created by taking a set of ideal key frames from my scenes in Blender, polishing them up a bit with new shaders and hand drawn adjustments, and then running these renders through a few different AI generators, such Seedance and Wan. Other tools used included EbSynth, Topaz and DaVinci Resolve.

I've stayed away from using AI in this manner, not out of any deep set principle, but because it sucks. Creative control over AI has been utterly useless. I would say that it's only starting to get competent, but has a ways to go.

With that in mind I realised that I've developed a fortunate position to test the creative utility of AI. My work already attempts to emulate hand drawn anime through CGI, but this has its limits because CGI is too perfect. It mathematically preserve 3D space, but real 2D art never does cause humans are flawed. But interestingly, if you supply a start frame and an end frame to an AI generator, and if those two frames contain instances of the same objects, and if the design of those objects differs ever so slightly, then the diffusion solver will smear those differences across the clip, so that you get that nauseating morphing effect that AI videos have been notorious for. But this doesn't happen if your objects are 3-dimensionally consistent, and this is what I wanted to test. Zoey is a 3D character, even though she's rendered like a drawing, that drawing still respects a consistent structure, and I was honestly curious whether or not AI would be able to latch on to that structure and create something passable.

There are still limitations with this approach. First off, it's expensive - this animation cost me around $400 to make. Every time you run a prompt you effectively roll a dice - you could get something good, could get something fuckn weird. I have a total of 172 clips that had been generated, and this resulting animation uses just 6 of them. Well, technically 6 and a half since I wanted the results from one fused with another and had to manually weave them together through photoshop and davinci. A lot of the effort to obtain desired results doesn't just come with luck but analysing and experimenting with the prompting and the references. There's a lot of pushing and pulling, seeing what works, what doesn't.

There was also a clip that had almost everything going for it but in it Zoey had 3 fingers instead of 4, and there was simply no way around that. Believe me I tried everything, all the way down to manually fixing the error over 250 frames by returning to Clip Studio Paint but the inconsistency looked disasterous. I find it hilarous that after a decade of an AI arms race, that all those parameters still can't figure out fingers.

Edit!: Apologies, on second reading the following is very poorly worded. It strongly suggests that I'm trading out blender, but that's not the case. This is a mixed bag of me sharing a general sentiment that I've always held, to produce more creative animation without losing so much time, which is a sentiment that doesnt have any necessary ties to AI. When I'm using blender, my goal is to find ways to speed up the process - that's a big thing that I've been doing for this year. Also the following is seeking feedback for how you guys feel about this form of animation and whether or not you'd like more of it in addition to blender.

Anyway, as jarring as it is, this is preferably how I'd like to make my animations, for numerous reasons. It's way, way faster. No gruesome inbetweening, no simulation or frame-by-frame sculpting. If I exclude the time taken to develop the key frames (the most essential component), then producing this animation took only 4 days. It's also exceptionally good quality.

Hopefully you guys feel okay about this direction. For me to make more content in this manner there's no less creative work than before. It just means I get to focus on the creative stuff. But let me know what you guys think.

Also quick note, this animation doesn't have Zoey's usual voice. The audio came with the generated clips, and I hadn't directed that ahead of time. Since I was just experimenting I didn't even expect to publish this. Basically I was quite satisfied with the sound effects it had supplied, but it also layered on top a standard anime voice which I couldn't separate out. Who knows, maybe there's an AI for that..

<3