Prompt Director

# Prompt Director Reference

Use this file when the user provides reference images, asks for a master prompt, needs face/color/logo replacement, uploads multiple images, or gives feedback that a generated result is wrong.

## Intent Priority

Pick the primary intent in this order:

1. **prompt-diagnosis** — the user says the output is wrong, too AI-looking, wrong color, wrong angle, unreadable, or unlike the reference.
2. **reference-to-master-prompt** — the user provides an image and wants the same style, composition, or aesthetic as a reusable prompt.
3. **realistic-human** — the request includes a face, avatar, UGC creator, selfie, portrait, or realistic person.
4. **brand-adaptation** — change brand colors, add logo, or match a brand guideline.
5. **video/storyboard** — motion, camera movement, shot list, or scene sequence.
6. **product/design-standard** — product shot, poster, social post, logo, infographic.

When multiple intents match, keep the highest-priority intent as primary and activate the others as submodules.

## Prompt Layers

Include these layers when useful, but only ask for missing layers that materially affect the result:

- **subject:** who or what is central.
- **identity_preservation:** face, hair, skin tone, body shape, outfit, silhouette, accessories to preserve.
- **environment:** setting, place, background.
- **composition:** crop, angle, framing, subject position, aspect ratio.
- **camera_lens:** smartphone, 35mm, 85mm, macro, wide, fisheye.
- **lighting:** window light, flash, studio, neon, cinematic, flat office.
- **color_palette:** dominant colors, brand colors, colors to replace.
- **style_mood:** realistic, UGC, editorial, premium, playful, futuristic.
- **typography_graphics:** headline, badge, logo, label, UI element; avoid long text inside AI images.
- **output_use:** Facebook post, Reels/TikTok, LinkedIn avatar, deck cover, landing hero.

## Reference Image Workflow

When the user uploads one image:

1. Identify the image role: **style/composition reference**, **face reference**, **logo/brand asset**, **product reference**, or **background/location reference**.
2. If the user is unclear, default to style/composition reference.
3. Write the master prompt using "the referenced subject" or "the provided face reference"; do not assign a real name unless provided by the user.
4. For face/color/logo replacement, explicitly say "use the uploaded face/logo/brand palette" instead of inventing details.
5. End with the exact assets still needed.

When the user uploads multiple images: