lesson 003: the ai skill nobody teaches you: intuition
Lesson three is about intuition with AI: knowing when to push a model, when to let go, and which one to reach for in each moment. To make that concrete I carry a single idea from a sketch all the way to a finished product film, an imaginary handheld for running Claude Code away from the desk, built end to end inside plnty. It runs through image generation with Uni 1.1 and GPT Image 2, variations and styler for the design, a 3D model and a Cinema 4D hero shot, a video restyle shootout across Kling, Ray 2, Aleph 2 and Seedance, then an upscale, a Figma mock-up and the final presentation.
What's inside:
Intro
0:00
ideation to execution, and what we're building: an imaginary device for claude code
The idea
0:31
a physical remote for running claude code when you step away from the desk
Inspiration + features
0:47
korg monotribe, tape decks and ti interfaces, plus the buttons it needs
Sketching the concept
1:35
dropping the sketch in, deleting what's wrong, and mapping the buttons
First frames: Uni 1.1 + GPT Image 2
2:07
two image models in batch mode, a panasonic look, and the splitter
refine
3:34
letting plnty rewrite a weak prompt into a detailed one for gpt image 2
variations
4:50
shape and proportion variations to pressure-test the design
styler
5:52
trying the device on in a range of styles
inpaint
6:24
eraser-mode cleanup on the edited image
close-ups
8:07
close-ups for the mechanical button, speaker and fine detail
Mock-ups + motion with Krea
9:13
dropping the device into scenes, fixing hands, and adding movement with krea
3D model + Seedance director
10:35
exploring angles and motion from a generated 3d model
Cinema 4D
12:58
judging the ai shot, then directing the hero shot by hand in cinema 4d
Video restyle: Kling, Ray 2, Aleph 2, Seedance
13:39
a four-way shootout for which model holds the text, then color and grain
Marketing materials + upscaling
15:24
extra imagery, with topaz vs seedvr for the upscale
Figma + presentation mode
17:12
into figma for a product page, then framing the final cut in presentation mode
The real lesson: intuition
18:01
when to push, when to let go, and which model to choose in each scenario
all
gen-ai
lessons
1Welcome to plnty
Plnty is open for Early Access. Co-founder Yambo walks through what we've been building: a three-dimensional workspace that holds image, 3D, and animation in one infinite, collaborative canvas. It's built for the part of the work where you don't know yet, where you've got nine half-ideas open and you're trying to see which one wants to grow.