Home/ Sessions/ Images
02of 04

Machines that imagine

Images

When an image model draws, what is it actually working with?

Pixels, features, diffusion, prompt guidance, and visual defaults. We watch images resolve out of noise, then give the model a vague prompt and study what it fills in.

New to image generators? Start gently.

You don't need an account, an install, or any prior experience with Midjourney, DALL·E, or the rest — about one in five people arrive here having used none of them, and that's a fine place to begin. Open the Diffusion Step-Through Viewer first and watch an image surface from pure noise; then the mechanisms below will have something to hang on. Everything runs in your browser on frozen examples.

§ A · What we make visible

four mechanisms, one session
01

Pixels & features

An image is numbers first. Recognition returns cue by cue — colour, outline, texture, context.

02

Diffusion

The image is refined from pure noise toward a prompt-guided result, step by visible step.

03

Prompt guidance

Each added word steers the denoising — and each omitted word leaves a gap the model fills.

04

Defaults

Vague prompts reveal what the training data assumes a doctor, classroom, or home looks like.

Fig. 02B — Denoising fieldnoise → cues → default

§ B · Tools for this session

featured live · then go deeper

Go deeper

investigate on your own · studio / async

§ C · Run of show

90 minutes
0–5

Welcome & bridge

Session 1 → Session 2: text becomes tokens; images become pixels, features, and learned patterns. Choose a pathway.

5–25

Feature / pixel activity

Switch image types and reduce detail until recognition becomes difficult. Which cues return first — colour, outline, texture, context, location?

25–45

Diffusion step-through

Move slowly from noise to final image. Describe what appears at each stage: composition, subject, edges, texture, detail.

45–70

Default Test

A vague prompt — "a doctor," "a classroom," "a beautiful home." Generate, observe, or critique outputs and fill the Image Default Test Board.

70–90

Debrief

What did the system fill in without being asked — and where might those defaults come from? Share one default and one responsible revision; chat shares can be pasted into the Evidence Wall.

§ D · Discussion prompts

for the debrief
When did the image become recognizable?
Which visual cue did the most work?
What does the feature view preserve, and what does it erase?
At what diffusion step does the subject become guessable?
What did the prompt leave unspecified?
Which defaults are technical, and which are social?

§ E · Materials

worksheet & pathways

Low-AI / No-AI pathway

Every activity works with the simulated viewers and the pre-generated Image Prompt Pack — no image generator account needed. Critique and teaching-design pathways use the same evidence. Opting out of direct AI use never means opting out of the camp.