> **Using this in ChatGPT web?** [Build a ChatGPT reference pack](NO1-ChatGPT-Pack.html). Select the actual face and finished style images, download together, then attach the extracted images and guide in your chat. This guide alone does not give ChatGPT access to the photos.
>
> [Download the complete library and skill](https://pub-a80d616c849c48839ddadd3e5c574946.r2.dev/NO1-Team-Brain-Setup.zip) — approximately 4 GB, including all locally available source photos, finished references, inspiration and instructions. Extract locally; do not upload this entire archive into one chat.

# NO1 Thumbnail Brain

Built from NO1 Creative and NO1 Admin Drive, plus your three PKGD reference collections and recent public YouTube thumbnails. Updated September 18, 2026.

The central rule is **a story moment with a clear unanswered question, rendered in the right client's visual language**. The consistent NO1 identity comes from composition, perspective, believable detail and emotional clarity. It does not require every video to use the same color grade, expression or text layout.

## What the brain contains

- 25 primary curiosity frameworks and 23 genre-specific methods from your operating system and course material.
- Client profiles for 4FUN, Groovy Gavin, Tyler Blanchard, Boneheads, Riley Coleman, Law Mother and 5Things, plus portfolio references for AMP, Airrack and Sticks.
- A visual library of 164 portfolio/collection files: 93 finished treatments and 71 sketches. Six sampled original shoot photos are also saved. Some files depict related or repeated designs.
- Hardscope has not launched its channel. Its 23 finals are explicitly confirmed by you, with separate guidance for Neon, Unpacked, Streamers 9–5 and Love & Justice.
- 577 published YouTube thumbnails across 12 launched channels, from March 18, 2025–September 18, 2026: 489 long-form videos and 88 stream replays. Shorts excluded. [Read archive coverage](no1-thumbnail-brain/references/youtube-archive.md). The original reviewed sample still supports the [client style guide](no1-thumbnail-brain/references/current-channel-styles.md).
- Your explicit rule: **Airrack and Airrack Uncut have the exact same thumbnail style.**
- A generation workflow that selects the moment, camera, light, references and quality checks for the video.
- A preference record for explicit feedback, so a correction can become a durable client rule rather than getting lost in a conversation.
- A final-quality standard drawn from 98 additional PKGD images, including six explicitly labeled box finals. The updated combined visual library contains **868 image files**, with stages kept separate.

[Browse the visual library](NO1-Visual-Library.html) · [Read the reusable skill](no1-thumbnail-brain/SKILL.md) · [View sources and coverage](no1-thumbnail-brain/references/source-register.md)

## The visual constants

**Freeze an event.** The frame should imply something happening: an action about to resolve, evidence of an unexplained event, an unexpected relationship, a meaningful choice. A generic pose is rarely enough. This is explicit in the [NO1 training manual](https://drive.google.com/file/d/1N0_iRU0WD2cDCZbbZJ47cQmKiz6Cq0FZ/view) and [department operating system](https://docs.google.com/document/d/1qnoR8la92Wmi-Eo4JVzLVS6N45VB_XivXI0H6Q9VJ8o/edit).

**Make perspective do useful work.** Your finished entertainment images often use a close face, hand or prop, then place the consequential action deeper in the scene. The camera creates involvement, scale or discovery. This is more specific than simply asking for a cinematic image.

**Build one coherent photograph.** Skin, clothing, objects and surroundings should share a believable light direction, perspective, contact and texture. A surreal premise can still have physically convincing execution. Crisp story-bearing detail matters; making every pixel equally sharp does not.

**Protect the hierarchy.** Your three-pane thinking gives each layer a job: attract attention, communicate the event, and add context or mystery. The layers should work together around one dominant relationship.

**Keep the promise honest.** The title, thumbnail and intro must work together. The image adds a visual question instead of copying the title word for word. These checks come directly from the [thumbnail checklist](https://docs.google.com/document/d/1Pe41cbynIFHlR2aOV5gdc8XMckiShjjTSUKPRDMFWC8/edit).

## How it adapts without losing the style

Your [Airack Uncut](https://www.pkgd.ai/share/review-folder/RYMI03VFNCYazZJH) and [Airrack](https://www.pkgd.ai/share/review-folder/KSmdUZoIzNXHjD5A) links set the final finish bar: natural skin and hair detail, coherent camera geometry, realistic contact and reflections, and rich environments with one clear visual hierarchy. They also show that restrained expressions and everyday lighting belong in the style. The [saved final-quality standard](no1-thumbnail-brain/references/final-quality-standard.md) identifies the exact references to load during generation.

| Video/client | Preserve | Change with the concept |
|---|---|---|
| 4FUN social comedy | Bright photographic finish, natural skin, expressive social relationships | Rule, opposing character, absurdity, crowd arrangement |
| Groovy Gavin suspense | Flash-lit subject, restrained concern, threatening depth | Unseen presence, evidence, setting, eyeline |
| Tyler experiments | Foreground test-object scale, force direction, tangible materials | Tool, environment, outcome question, test constraint |
| Boneheads survival | Close perspective and realistic body/prop contact | Scarce resource, improvised object, hostile environment |
| Riley education | Dark monitor setting, green/yellow hierarchy, legible charts/numbers | Actual concept, supported figures, progression |
| Law Mother explanation | Reassuring creator, clean diagram, concise label | Legal topic, meaningful object, explanatory relationship |
| 5Things interviews | Guest/host chemistry and recognizable store context | Gesture, prop, guest interaction, concise contextual text |

There is no verified global NO1 font or hex palette in the reviewed material. Imposing one would erase distinctions already present in your work. The brain stores palette roles and image references at the client level.

## What the sketch/final pairs teach

4FUN's laugh challenge preserves the foreground expression and background performer while improving texture, lighting and realism. Law Mother's house/keys and hierarchy examples preserve a simple explanatory arrangement. Riley's monitor layouts preserve the visual argument, but at least one sketch/final pair changes the numerical claim—so numbers must come from the current brief. Gavin's pairs preserve the near witness and distant threat, often adding direct-flash texture and atmosphere. Tyler's pairs preserve tool direction and oversized test-prop framing.

These pairings are visual correspondences, not an audited approval history. They teach how to preserve a concept during rendering without assuming every sketch detail is a permanent requirement.

## How future requests work

You can say: **“Use my NO1 thumbnail brain. This is a [video type] about [premise], for [client]. Make [concept].”** Add a working title, face/pose photo or approved sketch when available.

The brain identifies the right client treatment, chooses a framework, loads a few matching visual references, and generates the image. When you already know the concept, it proceeds to the image. When you want ideation, it proposes three different story directions with the same visual finish. Narrow revisions preserve everything outside the requested change.

For continuity, say **“Remember this for future 4FUN thumbnails”** or name the relevant client when giving a standing preference. The saved preference file distinguishes those instructions from one-off edits and from my interpretations of the archive.

## Evidence and limits

27 documents/templates were fetched, including three copies of the same shot-list template. Thumbnail-related sections of the broader training modules and Course Breakdown were reviewed. All 143 images returned by the seven-client curated collection and 21 portfolio images were visually inspected; six raw photos were sampled from 171 listed assets.

This is a substantial, source-grounded working library, not an exhaustive audit of every production folder in both drives. The source register identifies the unreviewed archive edges. No CTR or A/B dataset was analyzed, so the references are style evidence, not a ranked list of proven winners. The initial Hardscope gap is now resolved by 23 finals you confirmed in PKGD. Recent YouTube examples expand Sabri, Law Mother, AMP and Sticks; your personal favorite/rejected selections remain pending.

The reusable skill provides durable reference files for Codex on this computer. It is not model fine-tuning and does not automatically transfer into unrelated ChatGPT chats or synchronize with Drive.


## Pinterest inspiration

The [Inspiration library](NO1-Inspiration-Library.html) adds 6,169 distinct images from 6,491 collected pins. Search subjects, settings, actions, composition, mood, or lighting; board names are excluded from categorization. Ask: “Use the brain to find snow/helicopter inspiration for an Airrack concept.” The brain retrieves relevant images and adapts the idea to the client’s established finish. Initial tags are mostly source-derived and are visually checked when used. This bounded snapshot was finalized at your request; additional Pinterest pins remain outside it.