Inspiration
I wanted to make something without any dialogue with AI for the first time; a sort of challenge in visual storytelling for myself. I was also inspired by my own chameleon and I wanted to make something fun to put him in. I also had this thought put into my head from friends and colleagues that would ask me, "Why don't you just film the content you're making with AI?", and I thought about that for a moment... Why don't I? Well, there's the cost factors of course, but even if I tore that limitation away, would I still be using generative AI? So I thought I should try to make something that would be impossible (if not just really difficult) to make without AI, almost as an experiment. And this is what I came up with.
After experimenting with all of the different models, I was really impressed with how Google Veo 2 represented chameleons -- like the way it would make them move very slowly and cautiously. Every other model (including even the latest Veo 3.1) doesn't quite know the difference between chameleons and iguanas for some reason. Chameleons have zygodactyl feet with 5 fingers/toes, but every AI model seems to want to give them these reptilian 3-fingered toes! Nerdy, I know, but this excited me that Veo was getting it right! I wanted to see how far I could take it.
I was also experimenting with actually completely writing my own short through a sort of stream of consciousness generating. I usually have a whole script written out with the help of ChatGPT, Claude, Gemini, or any mix of those, but this time, I just decided to start generating and see where it would take me! I figured I would follow what I learned with my previous short films -- mainly, that character comes before everything -- and I followed my gut on what felt fun, cool, or funny and I think it turned out pretty fun to say the least.
I started this earlier in 2025, but I never ended up following through with it and it ended up being a project in a cabinet covered in cobwebs. I was getting stuck in my story and I didn't know what to do. I wanted the chameleon to come into the convenience store and stop a robbery, but all of the human interaction and dialogue just wasn't feeling great. So when The Chroma Awards popped up, I decided to pick this back up since I'd already done quite a bit of work on it. I showed my cut to my friend that's not in the GenAI world and they told me the people were freaking them out, and that I should just take the people out. Wow! What a revelation I had! At that point, I had already seen Flow (the cat movie) and I knew what I had to do!
I really had a lot of fun playing with the situational comedy in this. I was trying too hard before, but it really just works inherently as you have these animals that are out of place and you can really have a lot of fun with that!
Technically speaking, like I said, I mainly used Google's Flow platform. By the time I picked this back up, they had implemented Nano Banana and Imagen 4 natively into their system, so I was really streamlining things. I was a bit upset that Veo 3.1 had ruined my further chameleon generations, but I was happy that they decided to keep Veo 2 around; otherwise, I don't know what I would do. For both video and image, I had backups in case Google wasn't giving me what I needed. For images, I would edit things in Seedream v4 via fal (I still didn't run out of the $30 credit from this contest, that really came in handy!), and for any video generation that wasn't working, I threw into Kling 2.5 Turbo. It seems Kling is incredibly good with prompt coherence, but sometimes it just doesn't get anatomy or physics right where I need it to.
A really big part of the job was sound as well! I knew I didn't want to go too heavy-handed with music like I usually do and I wanted to challenge myself to see how far I could just take the soundscape. I initially took the challenge of trying to generate every sound in ElevenLabs, and it was working really well! Their generative sound effects platform is actually pretty cool and something I've been wanting for a while; however, after listening to this edit multiple times, I found that the quality of the generations was really bad when they were the only thing in the track. Especially if I had a long track with ambience, since you have a bit of low-quality sound in your ears for a while. For smaller effects, for the majority of all the fly buzzes, ElevenLab works quite well. So I would say I perhaps used Artlist.io for 75% of the soundscape and ElevenLabs for the other 25%.
I was looking forward to making my own music with Udio as well, but it would've taken me a lot more time and effort than I had for this deadline in order to make that happen, even for just three tracks. So I just stuck with my tried and true, Artlist.io for this as well.
Io, my panther chameleon, sadly passed away over the summer before I was able to complete this film.
I'd been avoiding this for some time actually, because I wasn't sure how I could handle so much time with a chameleon again after losing my buddy I would see every day... But I'm really proud with how this film came out and it was really nice spending so much time with my generated buddy again. I think it's perhaps helped grieve him in some way, so if anything, I'm happy I made this regardless of if it's any good or not. Because I know I will watch this and get some joy out of it and that's really all that matters to me with making art these days.
Log in or sign up for Devpost to join the conversation.