Experimenting with AI video

Cartoon test clip of a woman looking into a talking fridge, with a strip of frames and the cost of the generation

I’m experimenting with AI video generation and oh my god it’s wild.

It’s both incredibly advanced and I feel sorry at how much this is disrupting maybe an entire generation of artists.

At the same time, it’s still a lot of hard work to produce something that feels remotely close to “being right”. And I’m talking about a ~1 min video!

I didn’t want to be attached to a membership / specific tool so I went on with:

Opus directing
Nano Banana for initial image generation
Seedance 2.0 for video generation
Hyperframes for application frames
Elevenlabs for sound

It makes it feel specially high risk when the only thing you’re loosing is not just your time but a bunch of money in tokens!

I’m having a lot of fun though and looking forward to sharing the final result!

Originally posted on X