# Real-Life Premium Styleframes for Hybrid Motion Shorts

Use this when premium AI styleframes are visually better than deterministic layouts but still feel too abstract, like SaaS infographic symbolism rather than a credible everyday work scene.

## Core lesson

For TrueTraceShorts-style mechanism videos, premium quality alone is not enough. The viewer should immediately recognize a real work situation — meeting room, laptop, tablet, notes, project table, whiteboard — before reading the motion-graphic mechanism.

Preferred pattern:

```text
composition blueprint → real-life anchored AI styleframes → user selects direction → matching endframe → short I2V proof → full final only after proof acceptance
```

## Responsibility split

- **AI image generation owns:** realistic work-scene basis, premium product-film lighting, depth, atmosphere, text-free glass/holographic mechanism shapes.
- **Deterministic renderer owns:** all readable labels, arrows, captions, facts, stamps, callouts, sorting paths, and final assembly.
- **Wan/I2V owns:** subtle motion from accepted styleframes/keyframes only after user approval.
- **Never let AI own:** readable meeting notes, fake UI, factual labels, random abstract funnels, or the actual explanatory text.

## Approval gate

Only run this after a revision/approval like:

```text
REVISE_PREMIUM_STYLEFRAME_DIRECTION <candidate_id> ALL "...real-life/work-scene basis..."
```

Allowed:

- max 3 AI styleframes;
- contact sheet;
- delivery manifest with hashes and Telegram metadata;
- optional **endframe plan only** if one variant clearly wins.

Still blocked until explicit user approval:

- Wan/I2V;
- T2V;
- actual endframe image generation;
- full-final render;
- posting/platform APIs;
- render series.

## Recommended variants

Use real-life contexts rather than abstract diagrams:

1. **A — Desk / Laptop Scene**
   - realistic modern desk;
   - laptop/tablet with blank non-readable surface;
   - blank note cards;
   - subtle holographic information cards moving toward a generic summary area;
   - unconnected decision-target node upper-right.

2. **B — Meeting Room / Glass Board**
   - empty modern meeting room;
   - glass wall/whiteboard with blank post-it shapes;
   - transparent information flow into summary capsule;
   - separate unconnected target circle;
   - usually safest for caption-safe area and production clarity.

3. **C — Project Table / Work Evidence**
   - real project table with tablet, blank papers, blank note cards;
   - premium workflow visualization over the table;
   - strongest visual evidence feel, but higher risk of clutter/pseudo-writing.

## Prompt requirements

Each prompt should include:

- vertical 9:16;
- real-life work-scene basis;
- no people, hands, faces, silhouettes;
- no readable text, letters, numbers, fake UI, logos, brands, watermarks;
- blank papers/cards/screens only;
- subtle glassmorphism/holographic mechanism overlay;
- visible but unconnected decision target;
- bottom caption-safe area;
- renderer-owned overlay note.

Reusable negative prompt:

```text
readable text, letters, numbers, fake UI, logos, brand marks, watermarks, captions, subtitles, document text, charts with labels, people, hands, faces, human silhouettes, clutter, messy office, dark sci-fi tunnel, toy miniature look, PowerPoint diagram, low quality, blurry, random abstract shape, stock photo look, corporate cliché
```

## Quality gate

Auto-reject or revise if:

- the image is again purely abstract;
- it looks like a stock photo without a mechanism;
- readable text/pseudo-writing dominates;
- people, hands, faces, or silhouettes appear;
- logos/brands/UI appear;
- decision target is missing or connected to the flow;
- bottom caption-safe area is missing;
- the look is cheap corporate-stock, PowerPoint, sci-fi tunnel, or too busy.

## Evaluation pattern

For each variant report:

- real-life anchor strength;
- people/hands/faces: present/absent;
- readable text/logos/UI: present/absent;
- mechanism clarity;
- decision target separation;
- caption-safe area;
- I2V risk;
- own recommendation.

## Endframe planning without generation

If one variant clearly wins, create a text-only endframe plan, not an image. The plan should preserve:

- same camera/framing;
- same room/desk/table context;
- same lighting/materials;
- blank screens/cards/papers;
- no people/logos/text;
- decision target remains intentionally unconnected when illustrating the failure state;
- lower area remains caption-safe.

Then wait for:

```text
APPROVED_REAL_LIFE_STYLEFRAME_DIRECTION <candidate_id> <A|B|C>
```

Only after that may you generate a matching endframe and a short 5–6s Wan/I2V proof. Still do not run a full final.