clankermon

AI is infinitely more fun when it lets you do something you admire but don't quite understand 

Been in my ai-augmented character design bag over the weekend and opened up the video/animation bag last night 

7 year old me is dancing 

0x1cdbb57b12f732cfb4dc06f690acef476485b2a5

Mini apps will eat the internet | Latest drop: /clankermon | Building : /miniapps /large /fctrivia /flappycaster /jeeves /source @degentip @playsource /jump

what tool do you use for AI augmentation?

Midjourney / stable diffusion for ideation until I get enough outputs that are directionally correct

Triage in figma/illustrator

Then I train flux models on specific detail or styles, like linework/pixel, character sheet male or female 

Then for video I used wan 2.1 model, first time & don't know much about video models or what a good workflow would be yet

Pretty much this:

One thing I forgot to mention is the way I chain the flux models

Something that's working really well for me with clankermon is I trained a pixel sprite model that can take pretty much any description/image ref and spit out a sprite in the exact style

Good sprites capture deep detail in the most compact format, it almost feels like json for visual data

Then I run that through a the latest version of the main characters model (retrained often based on latest outputs)

Then through different models that bring our details like line work

Flux is a godsent for consistency

Then video is new to me as of the last 2 days

product /builders-garden /farville | 
 organizing /just-frame-it | hosting a web3 hub in rome /urbe-eth /ethrome 🇮🇹🐺

https://limone.lol

oh cool, i'm kind of fine tuning my strategy with midjourney currently

Pretty much this:

One thing I forgot to mention is the way I chain the flux models

Something that's working really well for me with clankermon is I trained a pixel sprite model that can take pretty much any description/image ref and spit out a sprite in the exact style

Good sprites capture deep detail in the most compact format, it almost feels like json for visual data

Then I run that through a the latest version of the main characters model (retrained often based on latest outputs)

Then through different models that bring our details like line work

Flux is a godsent for consistency

Then video is new to me as of the last 2 days 

https://warpcast.com/matthewfox/0xfd40293e