1sh1ro/ai4-photographer

AI4 Photographer

Core Workflow

1. Gather inputs:

   • Model photos or description.
   • Reference images, moodboard, or Xiaohongshu/XHS links.
   • Shoot goal: portrait, fashion, e-commerce, editorial, brand campaign, social cover, lookbook, portfolio, or AI-only concept.
   • Constraints: location, wardrobe, makeup, props, lighting gear, lens/camera, budget, aspect ratio, image count, and deliverable language.

2. If the user provides Xiaohongshu/XHS links, read references/xhs-reference-intake.md and use an available browser tool to inspect the visible page content, images, captions, and comments when accessible.

3. Analyze the reference style. For visual-heavy requests, read references/style-analysis.md.

4. Build a shoot script. For real photography execution, read references/shoot-script.md and use assets/shot-list-template.md as the output structure when helpful.

5. If the user wants AI-generated images, image prompts, or same-model style transfer, read references/ai-image-generation.md. Use the available image generation/editing tool when the user asks for actual generated images; otherwise deliver prompts and art-direction notes.

6. End with practical next actions: what to shoot first, what to prepare, and any missing inputs that would materially improve the result.

Output Standards

• Write in Chinese by default unless the user asks otherwise.
• Be concrete and production-ready: name poses, camera angles, focal length ranges, lighting placement, background, wardrobe, makeup, props, composition, crop ratio, and retouching style.
• Translate inspiration into executable instructions instead of copying a photographer's exact work.
• When references conflict, identify the dominant style and offer 1-2 alternate directions.
• Separate real-shoot instructions from AI-generation prompts so the user can use either path.
• For model photos, describe flattering choices based on visible features without making sensitive inferences. Avoid body-shaming language.
• If the model may be a minor, keep styling age-appropriate and avoid sexualized concepts.

Browser And Link Handling

Use Codex browser capabilities for live links when available. If browser tooling is not loaded, search for an in-app browser or Playwright browser tool first. If Xiaohongshu blocks access, requires login, or hides images, ask the user to open/login manually or provide screenshots, then continue from visible evidence.

Do not claim to have seen inaccessible images. Mark uncertain observations as inferred from captions, thumbnails, or user-provided screenshots.

Deliverable Shapes

Choose the smallest useful deliverable:

• Quick direction: 3-5 shoot concepts with mood, pose, light, and AI prompt seeds.
• Full shoot script: concept board, shot list, lighting diagram in words, styling, prop list, schedule, and post-production.
• AI image pack: reference analysis, identity/style constraints, prompt set, negative prompts, aspect ratios, and iteration notes.
• XHS sample deconstruction: creator/post summary, recurring visual grammar, usable takeaways, and risks/limitations.

Safety And Rights

Use public or user-provided references only for analysis and transformation. Do not instruct the user to impersonate a living photographer's exact signature style; instead describe visual attributes such as lighting, color palette, composition, lens feel, styling, and emotional tone.

For AI generation from a model photo, treat the supplied photo as authorized by the user. If the request implies deception, non-consensual sexualization, explicit content, or age-ambiguous sexual content, refuse that portion and offer a safe editorial/fashion alternative.