An amazing video.
Thank you! I'm understandably proud of it but I wouldn't say it's amazing... more "a decent attempt."

I would be very curious how much of the video was scripted and how much of the dialogue was directly AI generated. Also, curious about your entire process of creation, what took the longest, etc.
My process will probably explain a lot. After reading some threads on this forum, I looked around for affordable, above-board gen-AI video platforms. Most of them are prohibitively-expensive (for me, anyway, I'm all about cost-per-unit-of-entertainment) so I landed on Google's platform.
At "Pro" subscription level (about $20 monthly) you get 3 "chat-based" video generations a day. I use those to make completely random stuff and see if anything neat comes out of it.
You also get access to Flow that lets you generate tons of images with nano banana before it limits you with weird error messages.
So the process is... generate some throw-away idea videos... take ideas from that and storyboard with nano banana in Flow... use Veo in Flow to generate video clips... edit sequentially. Sometimes I have to re-generate previous moments when an especially good idea pops in... I would be a real filmmaker's nightmare. More Ed Wood than James Cameron.
I am NOT
pre-writing these (and here's where I disappoint myself)... I have the seed of an idea and I prompt and refine until I get the "performance" I want (or "close enough")... I pull it down and drop it into Kdenlive to incorporate into the video as I'm building it. As things are set up, I do write them down to checklist-payoff and the most I can manage is the 3-act structure of setup, mind-control pivot, payoff.
However, the dialogue is 90% mine. The 10% usually manifests as things like Ms. Archer's nervous chuckle after apologizing for "being so disagreeable"... when that "take" was delivered I absolutely embraced the ad-lib and tried to replicate it... with mixed results. The "oooh" during her statement of regret wasn't what I imagined (but more than acceptable)... and I settled on giving the model space to hallucinate serendipity instead.
What takes longest... is feeling around blindly as the model's safety features block my idea for seemingly no reason and I either have to go back and re-design what I want or I happen across the right combination of words to mitigate its fickle tastes. For example, I want a landscaper in a bikini and it's getting rejected as "adult material"
right now. Meanwhile... bikini on a beach? Fine. But you can't do YARD WORK on the beach. And I'm not calling it Beach Work.
