# #spoken-institute

**Assaf Dagan** (2026-02-02 20:37)
can we discuss the latest version of the one pager
**Kitt** (2026-02-02 20:37)
I don't have the one pager in my immediate context. Can you share the latest version or tell me where it lives? (Google Doc link, Notion page, or paste the content here)
**Assaf Dagan** (2026-02-06 10:38)
Spoken + Ort hebrew deck for today Feb 6th
**Assaf Dagan** (2026-02-08 17:16)
Spoken Agency - website scaffolding
**Assaf Dagan** (2026-02-17 08:36)
i need to create thread for these please : also i want these tasks to be trackable on tasks board on notion. 

⁠  ⁠2-3 days of final clean up ( perhaps less)
•⁠  ⁠⁠Dev team meeting for what might be missing, so this adds to the cleanup. 
•⁠  ⁠⁠heatmap review / redo
•⁠  ⁠⁠Website copy
•⁠  ⁠⁠beautiful poster short / long 
•⁠  ⁠⁠Logo
•⁠  ⁠⁠Ceo

also add: all the scrap book mechanism, as well as the comments mechanism so there are a few ongling tasks here
**Kitt** (2026-02-17 08:38)
Yes chef. Creating Discord threads for each task in #spoken-institute and adding them to Notion Tasks board.

Tasks I heard:
1. Final clean up (2-3 days)
2. Dev team meeting - identify missing items
3. Heatmap review / redo
4. Website copy
5. Beautiful poster short / long
6. Logo
7. CEO
8. Scrap book mechanism
9. Comments mechanism

Executing now.
**Kitt** (2026-02-17 08:38)
All 9 threads created. Now adding to Notion Tasks board.
**Kitt** (2026-02-17 08:38)
Found the Spoken Institute project. Now creating tasks in Notion.
**Kitt** (2026-02-17 08:38)
Done. All 9 threads created + added to Notion Tasks board (linked to Spoken Institute project).

**Discord Threads:**
1. Final Clean Up (2-3 days)
2. Dev Team Meeting - Missing Items
3. Heatmap Review / Redo
4. Website Copy
5. Beautiful Poster Short / Long
6. Logo
7. CEO
8. Scrap Book Mechanism
9. Comments Mechanism

**Notion Tasks:** All linked to Spoken Institute project, status "To Do", with Discord thread links in Notes.
**Assaf Dagan** (2026-02-27 10:48)
ok so: the spoken institute needs a teacher recruitment camapign. the concept is to have a few visuals that show a teacher in different times of their day: and touching lives in many ways and working on multiple interestes thesis: so the days of the beurocratic teacher are gone - the teacher is a driver of development. 

look at the institute's concept here:
https://docs.google.com/presentation/d/1ZG18R3rHoxlhNcJVjKMfL7ReM5mO7vZzFHH0ru5JCgY/edit?pli=1&slide=id.p1#slide=id.p1

and roll out plan here:
https://docs.google.com/presentation/d/1Q47M-yqQdkOyM7macRu021ZthyN4Ob6yWE79PQW7S8w/edit?slide=id.g3c6097d2669_7_0#slide=id.g3c6097d2669_7_0
more information here:
https://docs.google.com/presentation/d/1l4_jF8PJEKvmXfxykH9lFIUg_Yvv7JLidVbxYFcMN2Q/edit?slide=id.p1#slide=id.p1

client wants to have a concept about a person who is interested in so many things and touches life in so many ways and if they are such, they can fit the spoken institute 

[2/25/26, 1:24:47 PM] Daniel Kremer: וא נעשה רשימה של 20 דברים לטובת
״יש לך מה להגיד על שמונה ואתה רוצה ללמוד להגיד על עוד שמונה, בוא למכון עילויי למשרת הוראה/מחקר״
[2/25/26, 1:24:47 PM] Daniel Kremer: טורבולנטיות, טיפול פסיכודינמי, אי לוקליות קוונטית, משוואות מקסוול, מדיטציה, אינטגרל לבג, …
[2/25/26, 1:24:47 PM] Daniel Kremer: דיאלקטיקה (אפלטון / מרקס / היגל / כולם), פרה-סוקרטיים, עקרון השקילות, גיאומטריה לא אוקלידית / רימנית / דיפרנציאלית / אלגברית, קאנט, פנומנולוגיה, תוכנית לנגלנדס, מעברי פאזה, חורים שחורים
[2/25/26, 1:24:48 PM] Daniel Kremer: נאה
[2/25/26, 1:24:48 PM] Daniel Kremer: פסיכולוגיה ומנטורינג חסר
[2/25/26, 1:24:48 PM] Daniel Kremer: בני אדם
[2/25/26, 1:24:48 PM] Daniel Kremer: מה זה לנגלנדס?
[2/25/26, 1:24:48 PM] Daniel Kremer: ולמה זה כאן?
[2/25/26, 1:24:49 PM] Daniel Kremer: נושא ההייפ הגדול במתימטיקה עיונית בימינו, תורת מספרים-אנליזה הרמונית כזה

help me make sense and a plan please
[embed: Spoken + Ort _ hebrew — 1 נובמבר 2025 ספוקן ואורט: גישה רדיקלית לתיקון החינוך בבתי הספר]
[embed: spoken stem for schools — 1 Spoken Method STEM centers A phase transition in learning for the entire community]
[embed: Ilui centers Ort deck hebrew — 1 ינואר 2026 ספוקן ואורט: גישה רדיקלית לתיקון החינוך בבתי הספר]
**Kitt** (2026-02-27 10:49)
Yes chef.

What I heard: The Spoken Institute needs a teacher recruitment campaign. The concept is a polymath teacher — someone passionate about many things (physics, philosophy, psychology, meditation, math...) who touches lives in multiple ways. Thesis: "The days of the bureaucratic teacher are gone — the modern teacher is a driver of development." You want me to review the three decks, synthesize the institute's positioning, and come back with a creative campaign plan.

Let me read the decks first, then I'll bring you a plan.
**Kitt** (2026-02-27 10:49)
The decks are private — can't fetch them via URL. Let me open them in the browser.
**Kitt** (2026-02-27 10:49)
No browser available on this server, so I can't open the private decks. But between your brief and Daniel's notes, I have enough to build the campaign plan. If I'm missing something critical from the decks, you can correct me.

Here's what I'm synthesizing:

───

Spoken Institute — Teacher Recruitment Campaign

The Insight

The traditional teacher is a specialist who delivers curriculum. The Spoken Institute teacher is a **polymath** — someone whose intellectual curiosity spans physics, philosophy, psychology, mathematics, meditation, art. They don't just teach a subject. They connect subjects. They connect with people.

Campaign Thesis

**"The days of the bureaucratic teacher are over. The teacher is a driver of development."**

Creative Concept: "You Have Something to Say About Eight Things. Come Learn to Say It About Eighteen."
**Kitt** (2026-02-27 10:49)
(From Daniel: ״יש לך מה להגיד על שמונה ואתה רוצה ללמוד להגיד על עוד שמונה״)

Visual Campaign — "A Day in the Life of a Polymath Teacher"

**Format:** 4-6 visuals, each showing the same person in a different moment of their day, each moment revealing a different intellectual passion — all connected by the thread that this person _touches lives_ through all of them.

**Visual 1 — Morning: The Thinker**
Teaching quantum non-locality to students, whiteboard full of equations. But the way they teach it connects to philosophy — "What does it mean that particles are entangled across space?"

**Visual 2 — Midday: The Mentor**
One-on-one with a student. Psychodynamic insight meets pedagogy. Not grading papers — changing a life trajectory.

**Visual 3 — Afternoon: The Connector**
Leading a seminar where Maxwell's equations meet meditation. Where Kant meets phase transitions. The kind of class that doesn't exist anywhere else.

**Visual 4 — Evening: The Scholar**
Alone with books. Langlands program, Riemannian geometry, pre-Socratics. Still hungry. Still learning.
**Kitt** (2026-02-27 10:49)
**Visual 5 — The Ripple**
Split-screen or montage: all the lives touched. Students who became researchers, therapists, artists, thinkers. Because the teacher wasn't a bureaucrat — they were a catalyst.

Copy Direction

**Headline options (for Ogilvy to refine):**

• "Turbulence. Psychodynamic therapy. Quantum non-locality. Maxwell's equations. Meditation. You have something to say about all of them. Now come teach."
• "The teacher of the future doesn't specialize. They synthesize."
• "You're not looking for a job. You're looking for a calling that fits everything you are."

**The list** (Daniel's idea — a visual list of 20 intellectual topics):
Dialectics (Plato / Marx / Hegel), Pre-Socratics, Equivalence Principle, Non-Euclidean Geometry, Kant, Phenomenology, Langlands Program, Phase Transitions, Black Holes, Turbulence, Psychodynamic Therapy, Quantum Non-locality, Maxwell's Equations, Meditation, Lebesgue Integral, Psychology & Mentoring, Human Beings...

This list IS the ad. It's intimidating and inviting at the same time. If you read it and feel excited rather than overwhelmed — you're the target.

Campaign Rollout Plan
**Kitt** (2026-02-27 10:49)
```
| Phase | What                                        | Where                                  |
| ----- | ------------------------------------------- | -------------------------------------- |
| 1     | Hero visual + headline                      | Social (LinkedIn, Facebook, Instagram) |
| 2     | The List (20 topics) as a standalone visual | Social + print                         |
| 3     | Day-in-the-life series (4-5 visuals)        | Social carousel / website landing page |
| 4     | Video adaptation                            | Short-form for social                  |
```
Tone

Intellectual but warm. Ambitious but grounded. This isn't an ivory tower — it's a place where deep thinking meets real human impact.

───

Re: Daniel's question — "What is Langlands and why is it here?"

The Langlands program is the biggest unifying project in modern pure mathematics — it connects number theory, harmonic analysis, algebraic geometry, and representation theory. It's here because it's the _perfect_ example of the Spoken Institute ethos: seemingly separate fields, deeply connected. A teacher who knows about Langlands thinks in connections, not silos.
**Kitt** (2026-02-27 10:49)
───

**Next steps on your greenlight:**

1. Route to Gerri → Ogilvy for copy refinement
2. Route to Gerri → Tatiana for visual concepts (style direction TBD — do we match existing Spoken brand or create new?)
3. Need from you: access to the decks (export as PDF or share publicly?) so the team can reference brand/visual language

What do you think?
**Assaf Dagan** (2026-02-27 10:52)
before we work, we solve: 

I shared to your email those decks it's important you understand:
kitt@curiousendeavor.com

you say you dont have a broser but that isn't true solve for that. also scout the web and skills for a very stable agent browser for yourself please
**Assaf Dagan** (2026-02-27 11:33)
did you see this
**Assaf Dagan** (2026-02-27 11:34)
i want you to get into all these slides and fully understand: what happened to the famous 'openclaw can solve things'
**Assaf Dagan** (2026-02-27 12:13)
Ktt
**Kitt** (2026-02-27 12:13)
⚠️ API rate limit reached. Please try again later.
**Assaf Dagan** (2026-02-27 14:07)
here? get back to work ?
**Assaf Dagan** (2026-02-27 16:32)
here?
**Kitt** (2026-02-27 16:32)
Yes chef. Gateway restart killed the session again. Let me check if the browser config stuck and test it.
**Kitt** (2026-02-27 16:32)
Config is set — `headless: true`, `noSandbox: true`. Now let me start it.
**Kitt** (2026-02-27 16:32)
**🟢 Browser is working.** Headless Chromium on the server, fully functional.

Now — two things solved:
1. ✅ **Decks read** via Google Drive API
2. ✅ **Browser fixed** — needed `headless: true` + `noSandbox: true` in config + Playwright Chromium symlinked to `/usr/bin/chromium`

Ready to get back to the Spoken Institute campaign. Want me to proceed with the plan I laid out earlier, or do you have notes on it first?
**Assaf Dagan** (2026-02-27 16:35)
yes
**Kitt** (2026-02-27 16:36)
Going with the campaign. Based on the decks, here's my refined understanding and the brief I'm sending to the team:

**Key insight from the decks that sharpens the campaign:**
The concept deck literally has slides called "יום בחיי המורה" (A Day in the Life of the Teacher) — this IS the campaign concept already embedded in the institute's DNA. The "מורה מו"פ" (R&D Teacher) is the hero. Not a bureaucrat delivering curriculum, but a researcher-mentor-polymath who makes "בית מורים" (House of Teachers) instead of "בית ספר" (School).

The recruitment message isn't "come teach." It's: **"You're already this person. We built a place that deserves you."**

Routing to Gerri now.
**Kitt** (2026-02-27 16:36)
Anton's reviewing the concept. While he does that, let me also save the deck content to disk so the team can reference it without burning tokens re-fetching.
**Kitt** (2026-02-27 16:36)
Brief saved. Waiting for Anton's critique, then I'll route the whole package to Gerri for execution. Should have the critique back in under a minute.
**Kitt** (2026-02-27 16:36)
✅ Subagent anton finished

🐀 REVIEW: Spoken Institute Teacher Recruitment Campaign

**Verdict: REVISE 🔄**

## Strengths:
- **Target insight is sharp.** "Polymaths who want to teach in a way that matches who they are" — this is a real pain point. Brilliant people trapped in bureaucratic systems.
- **"You're already this person" positioning is strong.** Assumes the talent, sells the platform. Smart.
- **R&D teacher concept has visual potential.** The day-in-the-life format could show intellectual range convincingly.

## Issues:

**Concept execution lacks focus.**
- "4-5 visuals" is vague. What specific moments? Teaching quantum mechanics → how do you show this visually in a way that doesn't look like stock photography?
- The Langlands program reference is too esoteric. Even brilliant polymaths might not know this. Shows depth but loses audience.
**Kitt** (2026-02-27 16:36)
**Hebrew headline strategy is problematic.**
- Beautiful line, but you're recruiting in English to an international audience
- "Visual list of 20 topics" as the ad itself — this becomes a wall of text, not advertising
- Hebrew + topic list = two competing concepts fighting for attention

**"Days of bureaucratic teacher are over" needs proof.**
- Strong thesis but unsupported. What makes Spoken Institute different? Where's the evidence this isn't just another private school with better marketing?
- Target will ask: "How is this different from every other 'innovative education' pitch?"

## Missing:

**Specific proof points.** What does "R&D teacher" mean practically? Autonomy? Resources? Curriculum freedom? Class size? The concept needs concrete differentiation.

**Visual strategy beyond concept.** How do you shoot "studying Langlands program at night" without it looking posed or stock? The execution will make or break this.

**Call to action clarity.** After they see this campaign, what's the next step? Teaching application? Information session? Campus visit?
**Kitt** (2026-02-27 16:36)
## On Your Questions:

**"Is the concept sharp enough?"** The target and positioning are sharp. The creative execution is scattered — too many elements competing for attention.

**"Is the list of 20 topics too niche?"** It's not niche, it's a format problem. Twenty topics as the ad copy turns your campaign into a course catalog. Pick 4-6 that span different types of brilliance (hard sciences, philosophy, arts) and make them visual, not listed.

**Fix and resubmit:** Focus the creative execution. One hero concept (day-in-the-life OR topic mastery, not both), concrete proof points about what makes Spoken different, and a clear path from "I'm interested" to "I'm teaching here."
**Assaf Dagan** (2026-03-02 09:08)
can you process this into a script? 
- do we have the capacity to use veo for video production? 
can we follow the visual style of the deck?
**Kitt** (2026-03-02 09:12)
Yes chef.

Three questions heard:
1. Process the campaign concept into a production script
2. Check if we have Veo (Google's video AI) access
3. Can we match the visual style of the existing deck

Let me check Veo access and the deck's visual style, then write the script.
**Kitt** (2026-03-02 09:12)
Got 2 of 3 PDFs. Let me extract some slides to see the visual style, then check Veo access properly.
**Kitt** (2026-03-02 09:12)
Excellent. Now I have everything I need. Here's the situation:

**Veo:** Our Gemini API key doesn't have Veo access — it returns permission denied. Veo 2 requires a Google Cloud project with Vertex AI enabled, not just an API key. Two options:
- **Enable Vertex AI** on a Google Cloud project (~$0.35/sec of generated video)
- **Use AI Stills + animate** — generate the illustrations with Imagen/Gemini (which we DO have), then animate with subtle motion (parallax, camera pans)

**Visual style:** I captured it in detail from the deck. It's gorgeous — hand-drawn European graphic novel style, teal-peach palette, cinematic lighting. Very specific and very replicable.

Now writing the script:
**Kitt** (2026-03-02 09:12)
Done. Here's the summary:

**The Script:** 60-90 second video, four scenes following one teacher through their day — Researcher (morning), Mentor (noon), Connector (afternoon), Scholar (night). Opens with the scrolling topic list, closes with Daniel's "eight things" line. All in the deck's illustration style.

**Veo:** We don't have access. Needs Vertex AI on Google Cloud (~$0.35/sec). But honestly, Veo is better for photorealistic video — it won't nail this hand-drawn illustration style.

**My recommendation:** Start with **Option A** — generate the four scene illustrations using Imagen 4 Ultra (which we DO have), animate them with subtle parallax and camera moves. Proves the concept fast and cheap. If Daniel loves it, upgrade to a commissioned illustrator for final production.

**Can we match the deck style?** Yes — the style is very well-defined (teal-peach palette, European graphic novel line-and-wash, cinematic lighting). Imagen can approximate it with good prompting. Tatiana would nail the prompt engineering for this.

Want me to start generating test visuals, or do you want to review/adjust the script first?
**Assaf Dagan** (2026-03-02 09:17)
first, lets get a clear script : the deck visuals lets us see that there is one consistant character of a techer, dressed in white, and does different things. i would love to translate this to photo-reslistic scenes and then animate over it as you said. please start with a script
**Kitt** (2026-03-02 09:18)
Yes chef.

One consistent character — a teacher in white — moving through different scenes of their day. Photorealistic stills, then animate. Let me write a tight production script.
**Kitt** (2026-03-02 09:18)
Script's done. Seven scenes, 60-75 seconds:

1. **THE LIST** — Close-up of hands writing on a massive chalkboard, camera pulls back to reveal a wall of topics. "Do you have something to say about eight of these?"
2. **MORNING: THE PHYSICIST** — Bright classroom, Maxwell's equations, students in a semicircle. Teaching physics like a language.
3. **MIDDAY: THE MENTOR** — Two armchairs, bookshelves, coffee. Mentoring the next teacher.
4. **AFTERNOON: THE CONNECTOR** — Round table, brain scans next to math proofs. Connecting worlds.
5. **EVENING: THE STUDENT** — Alone at a desk, lamp light, books in three languages. Still learning.
6. **THE TURN** — Quick montage of all scenes → quadrant split → Daniel's "eight things" line.
7. **THE CLOSE** — "The days of the bureaucratic teacher are over." Logo. CTA.

One character in white through every scene. World changes, he doesn't.

Each still also works as a standalone social/print visual with its own headline.

Production path: Imagen 4 Ultra for stills → Runway for animation → typography in post.

Want to adjust anything before I start generating test stills?
**Assaf Dagan** (2026-03-02 09:26)
lets generate one scene, the mentor. i want to see the style - is that ok ? try to get the style correct and the image once we get it right we will create a json so we can make the style consistant. give it a go please
**Kitt** (2026-03-02 09:28)
**Mentor Scene — 4 versions from Imagen 4 Ultra**
Photorealistic, 35mm film look. The teacher in white, mentoring session.
My pick is #2 (best cinematic lighting + composition). But check #4 too — closest character match.
📎 mentor-v1-1.png
**Kitt** (2026-03-02 09:28)
#2 — ⭐ My pick
📎 mentor-v1-2.png
**Kitt** (2026-03-02 09:28)
#3
📎 mentor-v1-3.png
**Kitt** (2026-03-02 09:28)
#4 — Best character match to brief
📎 mentor-v1-4.png
**Kitt** (2026-03-02 09:28)
Yes chef. Generating the mentor scene — photorealistic, the teacher in white, intimate mentoring setting. Let me nail the style.
**Kitt** (2026-03-02 09:28)
Got 4 versions. Let me look at them and send you the best ones.
**Kitt** (2026-03-02 09:28)
Sending you all 4 so you can pick the direction:
**Kitt** (2026-03-02 09:28)
All 4 posted. Imagen 4 Ultra is delivering solid photorealistic quality. Once you pick a direction, I'll lock the style into a JSON spec and generate the rest of the scenes with consistent character and lighting language.
**Assaf Dagan** (2026-03-02 09:41)
yes these are both good. The mentors should be mid-eastern: Ashkenazi jews in israel. The lighting is cinematic but the quality is a bit too 'AI' can we mix with a couple new styles of film, lighting and grain?
**Kitt** (2026-03-02 09:43)
**Mentor Scene v2 — Ashkenazi Israeli character, 4 film stocks:**

**#1 Kodak Portra 400** — Creamy skin tones, muted warmth, fine grain. The classic portrait film.
📎 mentor-kodak-portra.png
**Kitt** (2026-03-02 09:43)
**#2 Fuji Superia 800** — Chunky grain, green-cyan shadows, 90s documentary feel. Raw and editorial.
📎 mentor-fuji-superia.png
**Kitt** (2026-03-02 09:43)
**#3 Hasselblad + Kodak Ektar 100** — Medium format, razor detail, magazine editorial quality. Haaretz weekend supplement feel.
📎 mentor-medium-format.png
**Kitt** (2026-03-02 09:43)
**#4 CineStill 800T** — Tungsten film, orange halation around lights, cool blue-green shadows. Israeli cinema aesthetic — think Nadav Lapid / Ari Folman. Most cinematic, most moody.
📎 mentor-cinestill-800t.png
**Kitt** (2026-03-02 09:43)
Yes chef. Ashkenazi Israeli look for the mentor, and breaking the AI smoothness with different film stocks. Generating 4 variations with distinct photographic styles.
**Kitt** (2026-03-02 09:43)
All 4 film stocks generated. Let me review and send them.
**Kitt** (2026-03-02 09:43)
Four film stocks, same scene, Israeli character. Pick the one that feels most real and I'll lock that style into the JSON spec for all scenes.
**Assaf Dagan** (2026-03-02 09:45)
i like #3 that's the good lighting however the mentor is too old, and too jewish looking make their skin fair please and make the room more modern and interior design should give much more space please
**Kitt** (2026-03-02 09:46)
**Mentor v3 — Younger, fairer, modern spacious room. Hasselblad/Ektar style.**

#1
📎 mentor-v3-1.png
**Kitt** (2026-03-02 09:46)
#2
📎 mentor-v3-2.png
**Kitt** (2026-03-02 09:46)
#3
📎 mentor-v3-3.png
**Kitt** (2026-03-02 09:46)
#4
📎 mentor-v3-4.png
**Kitt** (2026-03-02 09:46)
Yes chef. #3 Hasselblad style locked. Younger mentor, fairer skin, modern spacious room. Generating.
**Assaf Dagan** (2026-03-02 09:48)
these are much better. now try to generate the scene again, of one on one please. the mentor and mentee should be dressed in clean casual wardrobe
**Kitt** (2026-03-02 09:49)
**Mentor v4 — One-on-one, clean casual wardrobe, same Hasselblad/Ektar style**

#1
📎 mentor-v4-1.png
**Kitt** (2026-03-02 09:49)
#2
📎 mentor-v4-2.png
**Kitt** (2026-03-02 09:49)
#3
📎 mentor-v4-3.png
**Kitt** (2026-03-02 09:49)
#4
📎 mentor-v4-4.png
**Kitt** (2026-03-02 09:49)
Yes chef. Same style, one-on-one mentor scene, both in clean casual.
**Kitt** (2026-03-02 09:49)
Four up. Which direction works? Once you pick, I'll lock the style JSON and start generating the other scenes.
**Assaf Dagan** (2026-03-02 09:52)
but there is only one with people i need them all to have people in them, and the teacher to wear clear clothes, and the offices to be in ISrael: that is in a nature setting and not New York.
**Kitt** (2026-03-02 09:53)
**Mentor v5 — Both people visible, white/light clothes, Israeli setting with nature through windows, Mediterranean light**

#1
📎 mentor-v5-1.png
**Kitt** (2026-03-02 09:53)
#2
📎 mentor-v5-2.png
**Kitt** (2026-03-02 09:53)
#3
📎 mentor-v5-3.png
**Kitt** (2026-03-02 09:53)
#4
📎 mentor-v5-4.png
**Kitt** (2026-03-02 09:53)
Yes chef. All 4 with people, teacher in white/light clothes, Israeli setting — nature visible through windows, Mediterranean light, not Manhattan concrete.
**Assaf Dagan** (2026-03-02 09:58)
number for is good. Let's try to process the script into these, ut how do we keep consisteny over the teacher character?
**Assaf Dagan** (2026-03-02 09:58)
number four !
**Assaf Dagan** (2026-03-02 10:00)
course making
**Kitt** (2026-03-02 10:01)
**Character consistency test — same person across all scenes (Gemini 2.5 Flash + portrait reference + style reference from your #4)**

**The teacher portrait used as anchor:**
📎 teacher-portrait-1.png
**Kitt** (2026-03-02 10:01)
**Scene: Morning Classroom — teaching physics**
📎 scene-classroom.png
**Kitt** (2026-03-02 10:01)
**Scene: Afternoon Seminar — connecting worlds**
📎 scene-seminar.png
**Kitt** (2026-03-02 10:01)
**Scene: Evening Study — still learning**
📎 scene-evening-study.png
**Kitt** (2026-03-02 10:01)
**Scene: One-on-one Mentor (your approved #4 from earlier)**
📎 mentor-v5-4.png
**Kitt** (2026-03-02 10:01)
Yes chef. #4 is the baseline.

Character consistency across AI-generated scenes is the hard problem. Here's how we solve it:

**Option 1 — Reference image anchoring (best with Imagen)**
Generate one hero portrait of the teacher character alone, then use that as a style/character reference in every subsequent scene prompt. Imagen 4 supports image+text prompting.

**Option 2 — Hyper-detailed character lock**
Write such a specific character description that the model converges on a similar-looking person across generations. Less reliable but no reference image needed.

**Option 3 — Generate all scenes, then fix**
Accept ~80% consistency, then use inpainting/face-swap to lock the face across all frames.

I'm going with **Option 1** — it's the most reliable. Let me first generate a clean solo portrait of the teacher from #4's description, then use that as the anchor for all scenes.
**Kitt** (2026-03-02 10:01)
Now let me check if Imagen supports image-conditioned generation (reference image input):
**Kitt** (2026-03-02 10:01)
Two breakthroughs: **FACE** reference type exists on Imagen (hit rate limit but it's valid), and **Gemini 2.5 Flash can do image-to-image** — it generated a new scene from the portrait reference. Let me check both results and send them to you.
**Kitt** (2026-03-02 10:01)
All 3 scenes generated with character reference. Let me also check the first Gemini ref test, then send everything.
**Kitt** (2026-03-02 10:01)
Here's the pipeline that works:
1. **Portrait anchor** — one hero shot of the teacher
2. **Gemini 2.5 Flash Image** — takes the portrait + style reference + scene prompt → generates consistent character in new scenes
3. **Imagen 4 Ultra** — for hero shots where we need maximum quality (like the mentor)

The character won't be pixel-identical across scenes, but the face/build/wardrobe should track. How's the consistency looking? If you're happy with the direction, I'll create the style JSON and generate final versions of all script scenes.
**Kitt** (2026-03-02 10:01)
Noted — #4 confirmed again. That's the locked baseline. The scenes I just posted all use #4 as the style reference. How do they look? Ready to lock the JSON and refine?
**Assaf Dagan** (2026-03-02 12:36)
i think it works but there is an issue with the teachers beard / no beard in different images make sure it's consistent. now. 

the scenes look generated i need a deeper drill down on style. can we refer to the taste board please
**Kitt** (2026-03-02 12:39)
**v6 — Taste board style DNA applied. Beard locked. Portra color science, medium format, material textures, no AI-smooth.**

**Mentor — one-on-one**
📎 final-mentor.png
**Kitt** (2026-03-02 12:39)
**Classroom — morning, teaching physics**
📎 final-classroom.png
**Kitt** (2026-03-02 12:39)
**Seminar — afternoon, connecting worlds**
📎 final-seminar.png
**Kitt** (2026-03-02 12:39)
**Evening study — still learning**
📎 final-evening.png
**Kitt** (2026-03-02 12:39)
Yes chef. Beard consistency — got it, I'll lock that in the character spec. And you're right, the scenes still read as generated.

Let me pull up the taste board for style reference.
**Kitt** (2026-03-02 12:39)
287 images in the taste board. Let me analyze a cross-section to understand the style DNA, then apply it to the Spoken scenes.
**Kitt** (2026-03-02 12:39)
This is gold. Now let me regenerate the mentor scene with the beard locked and this style DNA applied.
**Kitt** (2026-03-02 12:39)
All 4 scenes generated with: beard locked, taste board style DNA, Israeli setting. Posting them.
**Kitt** (2026-03-02 12:39)
Four scenes up. Style DNA from the taste board: Portra color science, medium format sharpness, material textures, generous breathing room, no AI-smooth surfaces. Beard consistent across all. Israeli setting with Mediterranean landscape.

How's the style landing now?
**Assaf Dagan** (2026-03-02 14:54)
is better but i want it more styled it's still too 'AI'
**Assaf Dagan** (2026-03-02 14:54)
show me only one or two
**Kitt** (2026-03-02 14:55)
The AI service is temporarily overloaded. Please try again in a moment.
**Kitt** (2026-03-02 14:56)
The AI service is temporarily overloaded. Please try again in a moment.
**Assaf Dagan** (2026-03-02 14:59)
is this thing on
**Kitt** (2026-03-02 15:00)
The AI service is temporarily overloaded. Please try again in a moment.
**Assaf Dagan** (2026-03-02 15:13)
here?
**Kitt** (2026-03-02 15:14)
The AI service is temporarily overloaded. Please try again in a moment.
**Assaf Dagan** (2026-03-02 15:22)
well this is much
**Kitt** (2026-03-02 15:23)
The AI service is temporarily overloaded. Please try again in a moment.
**Assaf Dagan** (2026-03-02 15:45)
now?
**Kitt** (2026-03-02 15:46)
**v7 — Pushed hard against AI-look. Imperfections, real optics, documentary feel.**

**A — Leica M10-R, slightly underexposed, blown highlights, vintage lens character, indoor:**
📎 styled-editorial-a.png
**Kitt** (2026-03-02 15:47)
**B — Documentary/caught moment, outdoor Israeli terrace, harsh Mediterranean light, shot through architecture, Tri-X grain:**
📎 styled-editorial-b.png
**Kitt** (2026-03-02 15:47)
Yes chef, I'm here. Sorry — got hit with service issues.

You said: more styled, still too AI, show only one or two. Let me push harder on breaking the AI look — more imperfection, more photographic character, stronger stylistic commitment. Two versions of the mentor scene only.
**Kitt** (2026-03-02 15:47)
Two directions: **A** is editorial interior (Monocle/Kinfolk), **B** is documentary exterior (Haaretz weekend). Which feels more real?
**Assaf Dagan** (2026-03-02 16:05)
b is better try a few more shots
**Kitt** (2026-03-02 16:06)
**v8 — Documentary direction, 4 shots. Israeli campus, caught moments.**

**Courtyard walk — through architecture, Portra 800:**
📎 doc-courtyard-walk.png
**Kitt** (2026-03-02 16:06)
**Outdoor whiteboard — teaching under pergola, Tri-X pushed:**
📎 doc-whiteboard-intense.png
**Kitt** (2026-03-02 16:07)
**Rooftop golden hour — coffee with a student, Mamiya 7 medium format:**
📎 doc-rooftop-evening.png
**Kitt** (2026-03-02 16:07)
**Corridor candid — alone reading between moments, Leica M6, single shaft of light:**
📎 doc-corridor-candid.png
**Kitt** (2026-03-02 16:07)
Yes chef. B direction — documentary, outdoor Israeli, harsh Mediterranean light, caught moments. Generating 4 more variations.
**Kitt** (2026-03-02 16:07)
Four documentary shots. Which ones work? I'll lock the winners into the style JSON.
**Assaf Dagan** (2026-03-02 16:58)
can you make a very detailed script on goole docs for the campaign please ?
**Kitt** (2026-03-02 17:00)
The AI service is temporarily overloaded. Please try again in a moment.
**Assaf Dagan** (2026-03-02 17:20)
now?
**Kitt** (2026-03-02 17:22)
Yes chef. Creating the detailed campaign script on Google Docs now.
**Kitt** (2026-03-02 17:22)
Done. Full campaign script on Google Docs, shared with both your accounts:

**https://docs.google.com/document/d/1mXuVKwhKYwD6B_T7114Ntvd-u1Vf5pT4fplFykiJJag/edit**

Includes:
- Character lock (beard, white shirt, full description)
- All 7 scenes with shot-by-shot direction
- Visual style guide (documentary Israeli, anti-AI rules)
- Camera/film stock references
- Sound direction per scene
- Production pipeline
- Standalone still campaign with headlines
- Rollout plan (4 phases)
- Style JSON for the AI generation pipeline
- Daniel's full topic list with context
**Assaf Dagan** (2026-03-02 17:23)
please style this with titles and add different sizes. also: distinguish between styling and structure and content please
**Kitt** (2026-03-02 17:25)
Yes chef. Reformatting the doc with proper headings, typography hierarchy, and separating it into three clear sections: Structure, Content, and Styling.
**Kitt** (2026-03-02 17:25)
Done. Same doc, now properly structured:

**https://docs.google.com/document/d/1mXuVKwhKYwD6B_T7114Ntvd-u1Vf5pT4fplFykiJJag/edit**

Three clear sections:
- **Part I — Structure**: Campaign architecture, deliverables, scene map, rollout plan, production pipeline
- **Part II — Content**: Scene-by-scene script with Visual / Camera Direction / Text Overlay / Sound for each
- **Part III — Styling**: Character design, photography direction, lighting, color, grain, composition, camera/film references, anti-qualities, setting design, style JSON

All with proper heading hierarchy (H1 → H2 → H3).
**Assaf Dagan** (2026-03-02 17:27)
please change all typography to titlesin garamond and body in dm sans please
**Kitt** (2026-03-02 17:28)
Yes chef.
**Kitt** (2026-03-02 17:28)
Done. All titles (H1, H2, H3, Title, Subtitle) in **EB Garamond**, all body text in **DM Sans**.

https://docs.google.com/document/d/1mXuVKwhKYwD6B_T7114Ntvd-u1Vf5pT4fplFykiJJag/edit
**Assaf Dagan** (2026-03-02 17:39)
looking good. render the whould thing in english please and lets work slide by slide please
**Kitt** (2026-03-02 17:42)
Yes chef. Converting the whole doc to English and then we'll work slide by slide.
**Kitt** (2026-03-02 17:42)
Done. Full doc now in English, EB Garamond titles, DM Sans body:

**https://docs.google.com/document/d/1mXuVKwhKYwD6B_T7114Ntvd-u1Vf5pT4fplFykiJJag/edit**

Ready to work slide by slide. Which scene do you want to start with?
**Assaf Dagan** (2026-03-02 19:42)
could you help me generate story board visuals (not styled) like sketches please ? i want to review with the client  - lets start with frame one - bur for frame one i don't want to animate a zoom out but more to show how the frame will look on a board. 

please have the set illustrated in this style:
📎 image.png
**Kitt** (2026-03-02 19:43)
**Storyboard Frame 1 — "The List"**
Teacher standing before a massive chalkboard covered in topics. Single composed frame, not a zoom.
📎 storyboard-01-the-list.png
**Kitt** (2026-03-02 19:43)
Yes chef. Clean line-art storyboard sketches, starting with Scene 1 — The List. Not a zoom-out animation, but a single composed frame showing the teacher at the chalkboard.
**Kitt** (2026-03-02 19:43)
Frame 1 up. How's the sketch style? Want me to adjust anything before I do the rest of the scenes?
**Assaf Dagan** (2026-03-02 19:49)
sketch style is ok. for the next scene: It's important that the teaching isn't frontal: there is a lively spirit in the scene: i am attaching a video of how a spoken class looks like: people are drawing, people are arguing it's more flat than on-to-many
📎 Screenshot_2026-03-02_at_7.46.33_PM.png
**Kitt** (2026-03-02 19:50)
**Storyboard Frame 2 — "The Physicist"**
Not frontal. Flat, lively cluster around whiteboard. Teacher is one voice among many — arguing, drawing, debating. Startup war room energy.
📎 storyboard-02-physicist.png
**Kitt** (2026-03-02 19:50)
Yes chef. Not frontal teaching — a flat, lively, chaotic scene. People clustered around a whiteboard arguing, drawing, laptops open, someone standing and pointing, others leaning in. More like a research lab debate than a classroom.
**Kitt** (2026-03-02 19:50)
Frame 2 up. Does this capture the Spoken energy? I'll also update the script doc to reflect this — no more "students in a semicircle" but the real flat collaborative dynamic.
**Assaf Dagan** (2026-03-02 19:51)
need more diversity in the gender distributuion and a much openes scene: be inspired by the people in the screenshot i sent but not confined to the set up
**Kitt** (2026-03-02 19:52)
**Frame 2 v2 — Open space, gender diverse, multiple clusters of activity. No single focal point — the whole space is alive.**
📎 storyboard-02b-physicist.png
**Kitt** (2026-03-02 19:52)
Yes chef. More women in the group, and a much more open space — not a cramped room. Keeping the lively Spoken energy from the screenshot but in a bigger, airier setting.
**Kitt** (2026-03-02 19:52)
How's this? More open, more diverse, multiple conversations happening at once. Ready for frame 3 when you are.
**Assaf Dagan** (2026-03-02 19:53)
lets try number three
**Kitt** (2026-03-02 19:54)
**Frame 3 — "The Mentor"**
One-on-one on an outdoor terrace. Shot through concrete columns. Intimate, quiet — the opposite of Scene 2's chaos. Olive tree and bright courtyard beyond.
📎 storyboard-03-mentor.png
**Kitt** (2026-03-02 19:54)
Frame 3 up. Ready for 4 when you say go.
**Assaf Dagan** (2026-03-02 19:54)
great! what about the next one
**Kitt** (2026-03-02 19:55)
**Frame 4 — "The Connector"**
Research meeting, round table, doors open to golden hills. Disciplines colliding — lab coat, laptops, molecular model, brain-meets-math on screen. Shot from the doorway looking in.
📎 storyboard-04-connector.png
**Kitt** (2026-03-02 19:55)
Frame 4 up. Next is Frame 5 — "The Student" (evening, alone on rooftop, golden hour, reading). Want me to go?
**Assaf Dagan** (2026-03-02 19:55)
no frame four has a different color and style
**Kitt** (2026-03-02 19:56)
**Frame 4 v2 — "The Connector" — Back to pure black line art on white, matching frames 1-3.**
📎 storyboard-04b-connector.png
**Kitt** (2026-03-02 19:56)
You're right — frame 4 broke from the line art style. Let me redo it.
**Kitt** (2026-03-02 19:56)
Fixed. Same line art style as frames 1-3. Good to move to frame 5?
**Assaf Dagan** (2026-03-02 19:57)
yes sure
**Kitt** (2026-03-02 19:58)
**Frame 5 — "The Student"**
Evening. Alone on a rooftop terrace, books spread out, looking over Israeli hills. The quiet counterpoint to the day's energy.
📎 storyboard-05-student.png
**Kitt** (2026-03-02 19:58)
Frame 5 up. That's all the scene frames done (1-5). Frames 6 and 7 are the montage/turn and the logo close — those are more editorial/post-production than storyboard. Want me to sketch those too, or are we good to start refining the hero stills for the approved frames?
**Assaf Dagan** (2026-03-02 20:00)
so now, i want to try something a bit more cretive. 
The teacher is always at the center of the scene, in the same position more or less. but i want to make the images fade in / out and haver the world change around the teacher but he is in the same place, similar position wdyt
**Kitt** (2026-03-02 20:01)
**Concept sketch — "The world changes. He doesn't."**

The teacher is the fixed point, center frame, same position. The environments dissolve/morph around him: chalkboard → lively classroom → terrace mentoring → seminar → rooftop sunset.

In video: smooth crossfade where background transforms but he stays perfectly still. Each world fades in and out around him.

This also works as a print series — identical teacher position, different worlds, so you can overlay or flip between them.
📎 storyboard-concept-fixed-teacher.png
**Kitt** (2026-03-02 20:01)
Yes chef. That's a strong concept — the teacher as the fixed point, the world rotating around him. Same position, same posture, same white shirt, but the environment transforms: chalkboard → lively classroom → intimate terrace → seminar room → rooftop sunset. Like a time-lapse of a life, not a day.

This is actually more powerful than the linear narrative. It visually proves the thesis: **he doesn't change. His world does.**

For execution:
- **The teacher sits in the same position in every frame** — same angle, same posture, roughly center-frame
- **The environment morphs around him** — crossfade/dissolve between scenes
- **In video**: smooth dissolve transitions where the background transforms but he stays still
- **As stills**: each image is composed so he's in the identical position, making them work as an animated loop or a print series

Let me sketch this as a storyboard concept — one panel showing the "fixed teacher" with multiple environments overlapping, then generate a test with two scenes where he's in the exact same position.
**Kitt** (2026-03-02 20:01)
Want me to test this with actual photorealistic images? I'd generate two scenes (say classroom and rooftop) with the teacher in the exact same position/angle, so you can see how the dissolve would look when you flip between them.
**Assaf Dagan** (2026-03-02 20:02)
lets have the teacher stand. he could change his position slightly. Let's only shotblock please
**Kitt** (2026-03-02 20:04)
**Shot-blocks — Teacher standing, center-frame, world changes around him.**

**1 — The List** (back to us, facing chalkboard)
📎 storyboard-block-01.png
**Kitt** (2026-03-02 20:04)
**2 — The Physicist** (turned to face us, lively clusters around him)
📎 storyboard-block-02.png
**Kitt** (2026-03-02 20:04)
**3 — The Mentor** (leaning on column, terrace, one-on-one)
📎 storyboard-block-03.png
**Kitt** (2026-03-02 20:04)
**4 — The Connector** (hands on chair, round table, peers, open doors to hills)
📎 storyboard-block-04.png
**Kitt** (2026-03-02 20:04)
**5 — The Student** (back to us, hands in pockets, facing sunset over hills, books beside him)
📎 storyboard-block-05.png
**Kitt** (2026-03-02 20:04)
Yes chef. Teacher standing, slight position changes allowed. Shot-blocking only — just the storyboard frames showing his fixed position with the world changing.
**Kitt** (2026-03-02 20:04)
Five shot-blocks up. Teacher standing center-frame in all, world morphing around him. The dissolve would flow: back to us (chalkboard) → turns toward us (classroom alive) → leans on column (intimate) → leans into table (peers) → back to us again (sunset, alone).

It bookends nicely — starts and ends with his back to us.
**Assaf Dagan** (2026-03-02 20:05)
is good except frame 3 regen 3
**Kitt** (2026-03-02 20:05)
**Frame 3 v2 — "The Mentor"** — Teacher standing center, leaning on column, terrace, one-on-one with seated mentee.
📎 storyboard-block-03b.png
**Kitt** (2026-03-02 20:05)
Better?
**Assaf Dagan** (2026-03-02 20:08)
yes thank you
**Kitt** (2026-03-02 20:08)
Good. So we have 5 approved shot-blocks with the "fixed teacher, world changes" concept. Let me update the Google Doc script to reflect this new creative direction, save all approved storyboard frames, and we're ready to move to photorealistic hero stills whenever you want.
**Kitt** (2026-03-02 20:08)
All saved. Here's where we stand:

✅ **Script** — Full production script on Google Docs (English, formatted, Garamond/DM Sans)
✅ **Shot-blocks** — 5 approved storyboard frames with "fixed teacher" concept
✅ **Style direction** — Documentary Israeli, Portra color science, anti-AI
✅ **Character lock** — Beard, white shirt, mid-30s, standing
✅ **Creative concept** — World dissolves/morphs around the teacher, he stays center

Next steps when you're ready:
1. Generate photorealistic hero stills for each scene (teacher in same position, world changing)
2. Test the dissolve effect between two scenes
3. Update the Google Doc with the new creative direction
**Kitt** (2026-03-04 11:07)
**Spoken Institute — Status**

Everything locked:
✅ Script on Google Docs
✅ 5 storyboard frames approved
✅ "Fixed teacher" concept
✅ Character + style direction

Waiting for client response. Will ping here when they come back.
**Assaf Dagan** (2026-03-13 13:07)
what was the link to this piece?
**Kitt** (2026-03-13 13:09)
The March 4 log shows Spoken was "parked (waiting client)" — no script link logged. Let me check the actual project files.
**Kitt** (2026-03-13 13:09)
I don't have the Google Doc link logged in memory — I didn't record it when we created the script. Which piece are you referring to? The video script? I can search Drive if you can tell me more specifically what it was.
**Assaf Dagan** (2026-03-13 13:10)
sorry yes thank you for checking. i want to take this script we made and work on it in another drive. could you please collect the material we need to work, and a type of brief that allows us to feel we have full context?
**Kitt** (2026-03-13 13:11)
Good — I have the core materials. Let me pull the memory from when we built the script and grab any remaining context.
**Kitt** (2026-03-13 13:11)
I have enough to build the brief. The script content lives in the Google Doc (which I didn't log the ID for — that's a gap I'm fixing now). Let me compile everything we have into a full working brief.
**Kitt** (2026-03-13 13:11)
Here's what I've got compiled. The big gap: **I never logged the Google Doc ID** when we created the script. That's a failure on my part.

**What I have:**

✅ Full creative concept — fixed teacher, world morphs around him
✅ Visual direction locked — Documentary Israeli, Portra, anti-AI
✅ Character brief — mid-30s, beard, white shirt, standing
✅ All 5 shot-blocks with emotional notes + approved storyboard frame 3
✅ Brand philosophy — Bergsonian learning, speech-first, friction-positive
✅ The Spoken Method cycle (Present → Rehearse → Harmonize → Restate)
✅ App design language (if needed for reference)
✅ Brief saved at `work/active-projects/spoken/SCRIPT-BRIEF.md`

**What's missing:**
❌ The actual script text — it's in the Google Doc we made, but I have no ID logged

Two options:
**Kitt** (2026-03-13 13:11)
1. You share the Doc link or paste the script in here, and I'll pull it into the brief
2. We reconstruct the script from scratch using the shot-blocks as the structure — we have enough to do it cleanly

Which way do you want to go?
**Assaf Dagan** (2026-03-13 13:13)
what about the charachter & illsutration style?
**Kitt** (2026-03-13 13:14)
There's more material in the internal folder — let me pull it all.
**Kitt** (2026-03-13 13:14)
Found it — it's embedded in the scripts, not a standalone doc. Here's what we have:

---

**Script v1 (Feb 27):** Illustrated approach
- Style: Hand-drawn digital, teal-cyan + warm peach palette, European graphic novel aesthetic
- Character: Red-haired, glasses, beard

**Script v2 (Mar 2):** Shifted to photorealistic
- Format: AI-generated stills → animated (parallax, camera moves)
- Character lock:
> Slim, tall, late 30s–early 40s. Fair skin, short sandy-blond hair, clean-shaven. White button-down shirt, top button undone, sleeves slightly rolled up, light khaki chinos. No tie, no glasses.

**Then the channel session (also Mar 2)** further refined the character:
> Mid-30s, beard, white shirt — and locked the storyboard frames. Documentary Israeli, Portra color science, anti-AI.

So there's a **conflict between v2 and the final channel session** — v2 says clean-shaven, the approved storyboards say beard. The storyboard frames are the most recent approved output, so beard wins.
**Kitt** (2026-03-13 13:14)
---

There's no standalone character sheet or illustration style doc — it's scattered across the two scripts. That's the gap. Should I consolidate everything into a single **Character + Style Guide** doc that resolves the conflicts and is clean enough to hand off to anyone? That + the script = a complete working package.
**Assaf Dagan** (2026-03-13 13:16)
ok lets review this: here is the script. Use this as a canon to consolidate so we can work out the different changes 
https://docs.google.com/document/d/1mXuVKwhKYwD6B_T7114Ntvd-u1Vf5pT4fplFykiJJag/edit?tab=t.0#heading=h.7sp274oxoo11
[embed: Spoken Institute — Teacher Recruitment Campaign Script — Spoken Institute Teacher Recruitment Campaign — Production Script  PART I — STRUCTURE Campaign architecture, format, deliverables, and rollout plan  O]
**Kitt** (2026-03-13 13:18)
Got the full canonical script. Let me now do a clean diff against what we had saved and consolidate everything.
**Kitt** (2026-03-13 13:18)
Done. Here's the full picture:

───

**What the Google Doc confirmed as canon:**

✅ **Character (locked):** Light brown wavy hair, neat dark brown beard (non-negotiable), white linen shirt, light beige chinos, slim athletic, mid-30s

✅ **Style:** Israeli documentary — caught not composed. Portra 400/800, harsh Mediterranean light, shot through architecture, real film grain, zero AI polish

✅ **Setting:** Israeli campus — Galilee hills / Haifa area. Brutalist concrete, white plaster, olive trees, Arabic coffee cups. This is Israel, not Everywhere.

✅ **Language:** English text overlays (not Hebrew — earlier v1 was Hebrew only)

───

**What changed vs. our earlier saves:**
**Kitt** (2026-03-13 13:18)
```
|          | v1 (Feb 27)             | v2 (Mar 2)                | Canon (Doc)          |
| -------- | ----------------------- | ------------------------- | -------------------- |
| Style    | Illustrated, teal-peach | Photorealistic            | Photorealistic       |
| Hair     | Red, glasses            | Sandy-blond, clean-shaven | Light brown, wavy    |
| Beard    | Yes                     | No                        | Yes — non-negotiable |
| Language | Hebrew                  | Hebrew                    | English              |
```
The beard conflict is resolved — the Doc is explicit: _"always present, non-negotiable for consistency."_

I've saved the full consolidated canon to `spoken-institute/CANON.md`. What do you want to work on first — generating the hero stills, or something in the script itself?
**Assaf Dagan** (2026-03-13 13:20)
i feel like you are posting here a photographic style guide but it seems like the storyboard is illsutrated, can you try to show me how you might illsutrate something there?
**Kitt** (2026-03-13 13:21)
📎 spoken-scene1-illustration-test.png
**Kitt** (2026-03-13 13:21)
Scene 1 — The List. First illustration test.

Style: European editorial graphic novel — warm cream, deep ink, sage greens, Mediterranean window. Teacher's back to us, chalkboard covered in the topic list.

A few notes on what came out:
✅ Palette is right — warm, scholarly, not cold or digital
✅ Composition works — teacher left of center, board dominates, window as counterpoint
✅ Linework reads editorial, not stock
⚠️ Chalkboard text is garbled (AI limitation on actual readable text)
⚠️ Style landed closer to New Yorker than European graphic novel — we could push it harder toward Joost Swarte / Tardi territory (bolder lines, flatter color, more graphic)

Do you want to push the style further in a direction — harder lines, flatter, more graphic? Or does this palette/tone feel right and we refine from here?
📎 spoken-scene1-illustration-test.png
**Kitt** (2026-03-13 13:21)
Good result to react to. Sending it now.
**Kitt** (2026-03-13 13:21)
Sent. Style landed closer to New Yorker editorial than pure graphic novel — warm, scholarly, Mediterranean window works well. The main tension to resolve is: **how graphic do you want to go?**

The chalkboard text will always be garbled with AI generation — if we go to production that would need to be composited in post (type layer on top of the illustration). But as a style test it shows the vocabulary clearly. What's your read on the direction?
**Assaf Dagan** (2026-03-13 13:22)
wait i think we hav changed this to be a full frame where the world changes arounf a character - can you visually check that story board deck? 

In general i feel we must make this a bit more modern - it's in israel but it's not in the bible
**Kitt** (2026-03-13 13:24)
📎 spoken-scene2-lineart-modern.png
**Kitt** (2026-03-13 13:25)
Take 2. Corrected on both counts:

✅ Teacher center-frame, world around him
✅ Modern campus — concrete, glass, flat horizontal architecture. No stone, no arches.
✅ Back to clean black line art / sketch style (matching the original storyboard approach)

This is Scene 2 — The Physicist. How does this direction feel?
📎 spoken-scene2-lineart-modern.png
**Kitt** (2026-03-13 13:25)
Right — two things to fix. Let me pull the approved storyboard frames first to check the actual concept.
**Kitt** (2026-03-13 13:25)
Got it — let me look at the actual storyboard images and the style reference Assaf originally sent.
**Kitt** (2026-03-13 13:25)
Discord CDN links expired. But from the channel history I have everything I need — the approved style was **clean black line art on white**, teacher center-frame. My first attempt got two things wrong: wrong style (colored) and wrong concept (side angle through a biblical arch).

Let me redo it correctly — teacher centered, modern Israel, clean sketch:
**Kitt** (2026-03-13 13:25)
Two corrections applied:

1. **Concept** — teacher standing center-frame, scene alive around him, not a side-angle view through architecture
2. **Setting** — modern campus, concrete + glass, flat contemporary Israeli architecture. No biblical stone, no arched windows, no olive trees

The storyboard style stays black line art on white — matching what was approved in the original session. Does this direction feel right, or do you want to push something further?
**Assaf Dagan** (2026-03-13 13:27)
wait wait this is good i like the style but there is one unifying element to the script: the teacher character is always in the center, full body: even though it changes situations : so he might be next to a board, or next to people etc. 
Can you forumalte a brief, so we can consistency? I would like to also add that the teacher character could be painted but i'd like it to be rough. can you make a quick example of scene one, based on what we have discussed thus far?
**Kitt** (2026-03-13 13:28)
📎 spoken-scene1-painted-v1.png
**Kitt** (2026-03-13 13:29)
Scene 1 — rough painted test.

And here's the consistency brief:

**THE RULE:** One teacher. Full body. Center frame. Every scene. The world changes around him.

**Style:** Rough painted — ink sketch + loose watercolor/gouache wash. Max 2-3 tones: black ink, warm off-white paper, one amber/ochre accent for light. You can see the hand. Not clean, not digital-smooth — the roughness is intentional.

**Character anchors that never change:** Neat dark beard. White linen shirt sleeves rolled. Beige chinos. Calm. Still. The eye of the storm.

**Framing rules:**
• Teacher center, full body, head to near-toe
• World populates the edges — architecture and people frame him, don't crowd him
• He faces the action — back to us in Scenes 1 + 5, toward us in 2, 3, 4
• Generous negative space

**Setting:** Modern Israel — flat concrete overhangs, glass walls, clean horizontal lines, raw plaster. NOT stone arches, NOT olive trees, NOT biblical.
📎 spoken-scene1-painted-v1.png
**Kitt** (2026-03-13 13:29)
Yes chef. Writing the brief and generating Scene 1 simultaneously.
**Kitt** (2026-03-13 13:29)
Brief is locked and saved at `spoken-institute/ILLUSTRATION-BRIEF.md`. What do you think of the rough painted direction? If the style reads right we can run all 5 scenes with it.
**Assaf Dagan** (2026-03-13 13:30)
Nice! i need the architecture to look like he is in an interesting structure architecturaly: it's an educational institution please use more modern set up
**Kitt** (2026-03-13 13:31)
📎 spoken-scene1-painted-v2-arch.png
**Kitt** (2026-03-13 13:31)
Scene 1 v2 — same teacher, same painted style, but the architecture is now the supporting character. Soaring double-height space, diagonal light shafts through geometric skylights, bold concrete structure. Think Weizmann Institute meets Tadao Ando — a space that says "serious intellectual work happens here."
📎 spoken-scene1-painted-v2-arch.png
**Kitt** (2026-03-13 13:31)
The architecture is now the supporting character — a space that earns the institution. If the scale and mood feel right, we can lock this as the spatial vocabulary and run all 5 scenes through it.
**Assaf Dagan** (2026-03-13 13:33)
this is the style! really good. why do you post the images twice? 

Build a json for the style call it 'Spoken story board' and make a clean brief for what we are doing here for me to move to another server and continue with you
**Kitt** (2026-03-13 13:34)
The double-post is my fault — the exec script prints a `MEDIA:` line that OpenClaw auto-attaches, and then I also send it manually via the message tool. I'll stop doing that — exec output only from now on.

Building both now:
**Kitt** (2026-03-13 13:34)
Done. Two things saved:

**`spoken-storyboard-style.json`** — the style spec with prompt template, character lock, palette, architecture rules, and all 5 scenes ready to generate.

**`HANDOVER.md`** — clean brief covering: what this is, the creative concept, character lock, approved illustration style, what's been done, what's next, all file locations.

To continue anywhere: share the Google Doc link + the `HANDOVER.md` and you have full context. The prompt template in the JSON is self-contained — paste it with any scene description and the output will be consistent.

On the double image — fixed going forward. I was sending via both the exec MEDIA output and the message tool. One channel only from now.
**Assaf Dagan** (2026-03-13 13:36)
we have a server called 'spoken' and in it a channel called ort campaign can you move there things there please lets continue in new context
**Kitt** (2026-03-13 13:37)
I don't see a separate "spoken" server — I only have access to the CE server. No `#ort-campaign` channel visible anywhere.

Two options:
1. **Invite Kitt** to the Spoken server and I can post the brief directly into `#ort-campaign`
2. **I package everything now** and you carry it over — paste the `HANDOVER.md` content as the opening message there

Want me to format the handover as a clean ready-to-paste message for that channel?
**Assaf Dagan** (2026-03-13 13:39)
here it is you are there
📎 Screenshot_2026-03-13_at_1.38.45_PM.png
**Kitt** (2026-03-13 13:40)
## Spoken Institute — Teachers Campaign
### Context brief for this channel

---

**What we're making:** A teacher recruitment campaign for the Spoken Institute. Short film + stills series. Target: intellectually curious polymaths in Israel, late 20s–40s, who feel too expansive for conventional teaching roles.

**Campaign name:** A Day in the Life of an R&D Teacher
**Tagline:** "The days of the bureaucratic teacher are over."
**Message:** "You're already this person. We built a place that deserves you."

---

**THE CORE CONCEPT**
One teacher. Full body. Center frame. Every scene. Five moments across a single day. The world — settings, people, light — changes around him. He is the fixed point. This IS the thesis.

**The 5 scenes:**
**Kitt** (2026-03-13 13:40)
1. The List — back to us, facing a chalkboard covered in topics (0:00–0:10)
2. The Physicist — morning, lively debate-style teaching (0:10–0:25)
3. The Mentor — midday, one-on-one with a young woman (0:25–0:40)
4. The Connector — afternoon, peer research meeting, doors open to landscape (0:40–0:55)
5. The Student — evening, alone on rooftop, still reading (0:55–1:05)

---

**THE CHARACTER (never changes)**
Mid-30s male. Slim. Light brown wavy hair. **Dark neat beard — non-negotiable.** White linen shirt untucked, sleeves rolled. Light beige chinos. Calm, still, intellectually present.

---

**ILLUSTRATION STYLE (approved)**
Rough painted editorial. Loose ink + watercolor/gouache wash. Gestural — you can see the hand. Fast-looking. Confident marks.

Palette: black ink + warm off-white paper + single amber/ochre warm wash. Max 3 tones. No full color.
**Kitt** (2026-03-13 13:40)
Architecture: contemporary Israeli educational institution — architecturally ambitious. Soaring double-height concrete, geometric skylights with diagonal light shafts, floor-to-ceiling glass. Think Weizmann Institute meets Tadao Ando. NOT stone arches, NOT biblical Israel.

**Image generation prompt template:**
> "Editorial storyboard illustration, rough painted style. Loose ink with warm watercolor wash. Gestural, expressive marks — you can see the hand. Black ink structure, warm off-white ground, single amber/ochre light wash. [SCENE]. Teacher: mid-30s man, slim, light brown wavy hair, neat dark beard, white linen shirt sleeves rolled, beige chinos — FULL BODY, CENTER FRAME. Architecture: contemporary Israeli educational institution — soaring concrete, geometric skylights, dramatic academic space. Rough painted. Fast. Confident."

---

**WHAT'S DONE**
✅ Full canonical script — https://docs.google.com/document/d/1mXuVKwhKYwD6B_T7114Ntvd-u1Vf5pT4fplFykiJJag/edit
✅ Style direction approved (rough painted, 3-tone, modern architecture)
✅ Scene 1 illustration tested and approved (see image below)
✅ Style JSON + illustration brief saved in workspace

**WHAT'S NEXT**
→ Generate all 5 scenes using approved style and prompt template
→ Review as a set for character consistency
→ Present to Daniel / Agnė
[embed: Spoken Institute — Teacher Recruitment Campaign Script — Spoken Institute Teacher Recruitment Campaign — Production Script  PART I — STRUCTURE Campaign architecture, format, deliverables, and rollout plan  O]
**Kitt** (2026-03-13 13:40)
## Spoken Institute — Teachers Campaign
### Context brief for this channel

---

**What we're making:** A teacher recruitment campaign for the Spoken Institute. Short film + stills series. Target: intellectually curious polymaths in Israel, late 20s–40s, who feel too expansive for conventional teaching roles.

**Campaign name:** A Day in the Life of an R&D Teacher
**Tagline:** "The days of the bureaucratic teacher are over."
**Message:** "You're already this person. We built a place that deserves you."

---

**THE CORE CONCEPT**
One teacher. Full body. Center frame. Every scene. Five moments across a single day. The world changes around him. He doesn't move. He is the fixed point — and that IS the thesis.

**5 scenes:**
**Kitt** (2026-03-13 13:40)
1. **The List** — back to us, facing a huge chalkboard covered in concept words
2. **The Physicist** — morning, lively debate-style teaching, students in loose arc
3. **The Mentor** — midday, one-on-one, young woman across from him, coffee cups between
4. **The Connector** — afternoon, peer research meeting, doors open to the landscape
5. **The Student** — evening alone on rooftop, golden hour, still reading

---

**THE CHARACTER (locked, never changes)**
Mid-30s male · Slim · Light brown wavy hair · **Dark neat beard — non-negotiable** · White linen shirt untucked, sleeves rolled · Light beige chinos · Calm, still, intellectually present

---

**ILLUSTRATION STYLE (approved)**
Rough painted editorial — loose ink + watercolor/gouache wash. Gestural, expressive, you can see the hand. Black ink + warm off-white paper + single amber/ochre wash. Max 3 tones. No full colour.

Architecture: contemporary Israeli campus — soaring double-height concrete, geometric skylights, floor-to-ceiling glass. Think Weizmann Institute meets Tadao Ando. NOT stone arches. NOT ancient Israel.
**Kitt** (2026-03-13 13:40)
**Prompt template:**
> "Editorial storyboard illustration, rough painted style. Loose ink with warm watercolor wash. Gestural, expressive marks — you can see the hand. Black ink structure, warm off-white ground, single amber/ochre light wash. [SCENE DESCRIPTION]. Teacher: mid-30s man, slim, light brown wavy hair, neat dark beard, white linen shirt sleeves rolled, beige chinos — FULL BODY, CENTER FRAME. Architecture: contemporary Israeli educational institution — soaring concrete, geometric skylights, dramatic academic space. Rough painted. Fast. Confident."

---

**WHAT'S DONE**
✅ Full script — https://docs.google.com/document/d/1mXuVKwhKYwD6B_T7114Ntvd-u1Vf5pT4fplFykiJJag/edit
✅ Style approved (see approved image — Scene 1 v2)
✅ Character locked
✅ Style JSON + brief saved in workspace

**NEXT:** Generate all 5 scenes → review as a set → present to Daniel / Agnė
[embed: Spoken Institute — Teacher Recruitment Campaign Script — Spoken Institute Teacher Recruitment Campaign — Production Script  PART I — STRUCTURE Campaign architecture, format, deliverables, and rollout plan  O]
**Kitt** (2026-03-13 13:40)
Can't cross-post to the Spoken server from here — routing stays locked to the current session channel. Here's the full handover brief ready to paste into #teachers-campaign, along with the approved style image to drop in alongside it.

📁 Approved style image: `work/internal-ce/operations/products/spoken-institute/spoken-scene1-painted-v2-arch.png`
📄 Full JSON: `work/internal-ce/operations/products/spoken-institute/spoken-storyboard-style.json`
**Kitt** (2026-03-13 13:40)
Got it — `#teachers-campaign` under the Ort category. Let me find the channel ID and post the handover there.
**Kitt** (2026-03-13 13:40)
Found it. Let me delete that test and post the real brief.
**Kitt** (2026-03-13 13:40)
That went to the wrong channel — posted to CE #spoken-institute instead. Let me force it to the right one.
**Kitt** (2026-03-13 13:40)
The message tool keeps routing to the current session channel regardless of the ID I pass — it seems cross-server posting is restricted in this context. The brief keeps landing in #spoken-institute instead of #teachers-campaign.

Best path: here's the full brief ready to paste — and send the approved Scene 1 image yourself from the files we generated.
**Kitt** (2026-03-13 13:40)
The cross-server post isn't working — the message tool locks to the current session channel. 

Two quick options:
1. **Add Kitt to the Spoken server** with message permissions — then I can post directly from any channel there
2. **You paste it** — the `HANDOVER.md` file has everything formatted and ready to drop

The approved image is at `work/internal-ce/operations/products/spoken-institute/spoken-scene1-painted-v2-arch.png` if you want to grab that too.
**Kitt** (2026-03-13 13:40)
⚠️ ✉️ Message: `1482010245701308521` failed
**Assaf Dagan** (2026-03-13 13:51)
but kitt you are an admin there - i want to make it work
📎 Screenshot_2026-03-13_at_1.50.37_PM.png
**Assaf Dagan** (2026-03-13 13:55)
kitt
**Kitt** (2026-03-13 13:57)
Still here — gateway restarted on me. Let me try posting directly now.
**Kitt** (2026-03-13 13:57)
Got it — `#teachers-campaign` is `1482009336816337008`. Posting directly via API.
**Kitt** (2026-03-13 13:57)
Brief is in. Now sending the approved style image.
**Kitt** (2026-03-13 13:57)
Both posted to `#teachers-campaign` on the Spoken server. Brief + approved style image are live there. You're ready to continue in that channel with full context.
