# Teacher's Guide — Session 8 🟢
# “Create with AI: your creative studio”

**Program:** Applied AI — Beginner Level (from 12 years old, general public)
**Instructor:** Yann Isola
**Recommended duration:** 2 hours (adaptable 1h30 – 2h30)
**Prerequisites:** Sessions 1 to 7 (concepts: AI, learning, prompting, generative AI, ethics, application sectors)

---

## 1. Educational objectives

At the end of the session, each participant should be able to:

1. **Explain** the difference between *using* a generative AI (S5) and *creating with* an AI: direction, iteration, choice.
2. **Apply** the 4-step method of the creative studio: **Idea → Prompt → Critique → Iteration** (I-P-C-I loop).
3. **Write** a complete creative prompt with the 5 ingredients: subject, style, atmosphere, details, constraints.
4. **Discuss** the question “who is the author of a work created with AI?” » citing at least 2 arguments on each side. ⚠
5. **State** the 3 rules of creative honesty: say when you have used AI, do not copy the style of a living artist without quoting them, check before publishing.
6. **Carry out** as a team a creative mini-project (the “client order”) by applying the I-P-C-I loop.

> ⚠ **Volatile content**: copyright on works generated by AI evolves (court decisions, national laws, conditions of use of tools). Check legal news **before each session** (15 min wake-up time). Passages marked ⚠ should be re-checked as a priority.

---

## 2. Overview and common thread

**Code of the session:** *“In session 5, you discovered that AI knows how to draw, compose, write. Today, we are changing roles: you are no longer spectators, you become artistic directors. AI is your studio — but YOU are the artist. »*

This is the “workshop” session of the program: 60% practice. It recycles and deepens three acquired knowledge:
- the **prompting** of S4 (we go from the utility prompt to the creative prompt),
- the **generative tools** of S5 (we move from discovery to method),
- the **ethics** of S6 (we apply it to a concrete case: copyright).

The highlight is the **final studio project**: each team receives a fictitious “client order” and must carry it out by applying the I-P-C-I loop, then present its result AND its process.

**Educational posture:** you are the “producer” of the studio: you give the orders, you time, you push to iterate (“that’s good… but can we do better?”). Value the *process* as much as the result: a team that iterated 4 times with an average result worked better than a team that stopped at the first successful attempt by chance.

### Timed course (based on 2 hours)

| Time | Sequence | Format |
|---|---|---|
| 0:00 – 0:08 | Tagline: “Who made this image? » | Discussion |
| 0:08 – 0:20 | Use vs. create: the role of the artistic director | Interactive presentation |
| 0:20 – 0:35 | The I-P-C-I loop: live demonstration ⚠ | Commented demo |
| 0:35 – 0:50 | The creative prompt: the 5 ingredients | Presentation + exercise 1 |
| 0:50 – 1:00 | ☕ Break | — |
| 1:00 – 1:15 | Who is the author? Copyright and AI ⚠ | Mini-framed debate |
| 1:15 – 1:25 | The 3 rules of creative honesty | Presentation + practical cases |
| 1:25 – 1:50 | 🎬 Studio project: “the customer’s order” | Workshop (exercise 3) |
| 1:50 – 2:00 | Lightning presentations + summary + teaser quiz | Restitution |**Adaptation 1h30:** reduce the demo to 10 mins, merge debate and rules of honesty (15 mins), studio project to 18 mins with 1-sentence presentations.
**Adaptation 2h30:** add the complete exercise 2 (the respectful remix) after the break, and extend the final presentations (2 min per team with questions).

---

## 3. Material

### Essential
- A4 sheets + colored markers (at least 4 colors per table).
- Printed “customer order” sheets (models in Appendix A of this guide).
- Whiteboard or flipchart to draw the I-P-C-I loop.
- The slideshow of the session (`slides/slides.md`) projected, or printed as a backup.

### Recommended
- A connected station for the live demo (an image OR text generation tool). ⚠ Test the tool **the day before**: accounts, quotas, content filters.
- The session web page (`webpage/index.html`) open on workstations or freely accessible (works offline, no account required).
- A visible timer to pace the workshop.

### 100% screen-free version
The entire session works without a machine: the I-P-C-I demo is done with **you as a human AI** (see §5, box “demo without screen”), and the exercises all have a paper version.

---

## 4. Sequence 1 — Hook: “Who made this image? » (0:00 – 0:08)

Project (or describe) an AI-generated image based on a highly crafted prompt — for example: *“a lighthouse atop a flying whale, 1930s travel poster style, golden evening light”*.

Ask the question: **“Who made this image?” »** Let the answers come: “AI”, “computer”, “person”…

Then reveal the prompt and ask: **“What now?” The idea of ​​the whale, the lighthouse, the 1930s style — where did it come from? »** Answer: from a human. The AI ​​executed; humans imagined, chose, directed.

> 💡 **Key message of the hook:** *AI without human ideas produces nothing interesting. Creativity is choice — and choice is human.*

⚠ Be careful not to decide too quickly: some participants will defend “AI is the artist”. Don’t close the debate — say “we’ll come back to it after the break, keep your arguments!” ". This feeds into the 1:00 mini-debate.

---

## 5. Sequence 2 — Use vs create + the I-P-C-I loop (0:08 – 0:35)

### 5.1 Use vs create (0:08 – 0:20)

Central analogy: **the movie director**. He doesn't act, doesn't film, doesn't compose the music — and yet it's HIS film. For what ? Because he **decides**: the scenario, the framing, the takes we keep, those we reshoot.

With generative AI, it’s the same:

| Passive user | Artistic Director |
|---|---|
| Type a vague request | Describes a precise vision |
| Take the first result | Compare, review, redo |
| Can't say why it's good/failed | Knows what he wants to change |
| Stops there | Iterates to HIS vision |

Question to the room: “When you ask an AI for an image and you take the first one, who created what? » — expected: the AI ​​chose everything (composition, colors, etc.), the human only made a trigger.

### 5.2 The I-P-C-I loop (0:20 – 0:35) ⚠

Draw the loop on the board and bring it to life live:

1. **Idea** — what do I want? (vision, emotion, use)
2. **Prompt** — I describe it precisely (the 5 ingredients, see sequence 3)
3. **Critical** — I look at the result: what works? what is missing?
4.**Iteration** — I modify the prompt (or change my mind) and start again.

**Live demo (if screen):** start with a deliberately poor prompt (“draw a dragon”), show the generic result, then iterate 3 times, commenting on each review out loud. The goal: to show that the 4th result is *yours*, not that of the machine. ⚠ Generate your backup examples **the day before** in case the tool is down or filters the request.

> 🎭 **Screenless demo:** play the AI ​​yourself! The participants give you a prompt, you draw on the board in 30 seconds, deliberately literal and clumsy (like an AI that takes everything literally). They criticize, reformulate, you redesign. Laughter guaranteed and the I-P-C-I loop is understood forever.

**Point of attention ⚠:** some participants will conclude “so we need very long prompts”. Correct: a good prompt is *precise*, not necessarily long. Too many contradictory constraints = confusing result.

---

## 6. Sequence 3 — The creative prompt: the 5 ingredients (0:35 – 0:50)

Express recall of S4 (utility prompting: role, context, task, format). Today we add the **artistic** dimension. The 5 ingredients of the creative prompt:

1. **Subject** — what? (“a fox astronaut”)
2. **Style** — like what? (“watercolor”, “pixel art”, “film photo”)
3. **Atmosphere** — what emotion? (“melancholy”, “joyful”, “disturbing”)
4. **Details** — what specifically? (“cracked helmet, flag stuck in a cheese”)
5. **Constraints** — what to avoid/impose? (“no text in image”, “poster format”)

Complete example to build WITH the room, ingredient by ingredient, on the board.

**Start exercise 1** (“The prompt in 5 ingredients”, cf. `exercises/exercises.md`): 12 min, in pairs, 100% paper. Corrected: a good exercise prompt contains the 5 identifiable ingredients and without contradiction. Have 2-3 prompts read out loud and have the rest of the room identify the ingredients.

**Typical answer key (exercise 1)**: *“A library under the ocean (subject), stained glass style (style), calm and mysterious atmosphere (ambience), with fish reading luminous books (details), without any visible human (constraint). »*

---

## 7. Sequence 4 — Who is the author? (1:00 – 1:15) ⚠

After the break, a structured mini-debate (“two camps + judges” format) on: **“An image generated by AI: who owns it? »**

Give the factual elements BEFORE the debate (3 min):

- In many countries, a work must come from a **human** to be protected by copyright. An image generated “alone” by an AI is often not protected for anyone. ⚠ (state of the law to be re-checked regularly: current case law in several countries)
- The more the human has **directed** the work (detailed prompts, iterations, retouches), the more they can claim a creative contribution.
- AIs have learned from **millions of existing works**, often without the artists' consent → this is the source of anger among many creators, and ongoing lawsuits. ⚠
- The conditions of use of the tools vary: some authorize the commercial use of the generated images, others not. ⚠

Arguments expected on the “human is the author” side: the idea, the direction, the iteration, the final choice.On the “it’s not that simple” side: the AI ​​made the concrete visual choices; it was nourished by the work of other artists; without them, nothing would exist.

**Point of attention ⚠:** do NOT give definitive legal advice (“you have the right to…”). Stay on: “the law evolves, in case of doubt for public or commercial use, we inquire”. This is the honest and only sustainable posture.

---

## 8. Sequence 5 — The 3 rules of creative honesty (1:15 – 1:25)

To write on the board and copy (this is THE summary to remember):

1. **I say it.** When I publish or submit work done with AI, I mention it. (Direct link with S6: transparency.)
2. **I respect artists.** I don't ask "in [living artist] style" to make it sound new, and I don't pass off someone's style as my own.
3. **I check before publishing.** Exact content? Not hurtful? No recognizable real person without agreement? (Link with S5 deepfakes.)

Quick practical cases (answer “ok / not ok / it depends”):
- Illustrate your school presentation with an AI image **by saying it** → ok.
- Win a drawing competition with an AI image **without saying it** → not ok (cheating).
- Generate “in the style of Van Gogh” → it depends: artist long dead, work in the public domain — acceptable, but it is mentioned.
- Generate “in the style of [living illustrator]” and sell the result → not ok.

**Expected correction:** the important thing is the reasoning (transparency, respect, verification), not the binary response.

---

## 9. Sequence 6 — Studio project: “the client’s order” (1:25 – 1:50)

This is exercise 3 of file `exercises/exercises.md`. Teams of 3-4. Each team draws an **order form** (appendix A): a fictitious client, a creative need, two constraints.

Unfolded on the instructor side:
- 2 min: drawing orders + reading.
- 5 min: **Idea** phase (brainstorm, intention sketch).
- 6 min: **Prompt** phase (writing with the 5 ingredients).
- 5 min: **Critical** phase — each team passes its prompt to the neighboring team, who “plays the AI” by drawing/describing the literal result, then returns it with 2 critiques.
- 5 min: **Iteration** phase (improved v2 prompt).
- 2 min: preparation of the restitution.

**With screens:** replace the “neighboring team plays the AI” phase with real generation, but keep the exchange of critiques between teams.

**Success criteria (corrected):** ① the 5 ingredients present in the v2 prompt; ② at least 2 clear differences between v1 and v2, justified by the criticisms received; ③ the customer's order respected (need + 2 constraints); ④ intended honesty (“created with the help of an AI”).

**Point of attention ⚠:** teams will want to “win” by mocking the drawing of the neighboring team. Remember the studio rule: *we criticize the prompt and the result, never the people.*

---

## 10. Sequence 7 — Restitution and synthesis (1:50 – 2:00)

Each team: 1 minute — the command, the v1 prompt, what the review changed, the v2 prompt. Systematic applause.Summary on the board (complete by the room):
- Generative AI is a **studio**, not an artist: it executes, the human directs.
- The method: **Idea → Prompt → Critique → Iteration**, in a loop.
- A creative prompt = **5 ingredients**: subject, style, atmosphere, details, constraints.
- The copyright of the AI ​​works is **under construction** ⚠; When in doubt, ask for information.
- The **3 rules of honesty**: I say it, I respect the artists, I check before publishing.

Quiz teaser (2 oral questions, see `quiz/quiz.md`) and announcement of session 9.

---

## 11. Appendix A — “Customer order” sheets (to print and cut out)

1. **The “Au Pain Levant” bakery** wants a poster for its new giant croissant. Constraints: warm atmosphere in the morning; no text in the image.
2. **The town hall** wants a mascot for waste sorting week. Constraints: non-human mascot; colors of nature.
3. **A middle school music group** wants a cover for its first title “Summer Storm”. Constraints: electrical atmosphere; without a recognizable face.
4. **The municipal library** wants a picture for its fantasy book club. Constraints: illustrated story style; an animal should appear.
5. **The astronomy club** wants a logo for its star night. Constraints: readable in small format; maximum 3 colors.
6. **An animal shelter** wants a thank you card for its donors. Constraints: mild emotion; no text.

---

## 12. Appendix B — Instructor FAQ

**“Will AI replace artists? »** — Link to the golden rule of S7: AI assists, it does not replace. Creative professions are evolving (artistic directors use these tools), but intention, taste and choice remain human. Be honest: certain markets (quick illustration, image bank) are really disrupted. ⚠

**“Is it cheating to use AI for homework?” »** — It depends on the rules set by the teacher: if it's forbidden, yes; if it is allowed and mentioned, no. Refer to rule 1 (I say it).

**“AI images, can we sell them? »** — “It depends on the tool and the country, and it changes quickly. Before any commercial use: read the conditions of the tool and obtain information. » ⚠ Do not improvise a legal response.

**“Why does the AI ​​miss hands/text in images? »** — It reproduces learned statistical patterns (S2, S5): details with high variability (fingers, letters) are the hardest to model. It is also a good detection index (deepfakes link, S5). ⚠ Models improve quickly: re-check if the example still holds.

---

*End of guide — Session 8 🟢 · Applied AI · Yann Isola. Next step: Session 9.*