guangjun5952/tvc-ai-director/tvc-director-treatment

TVC Director Treatment

Use this skill to translate an approved creative concept into a rich director treatment and visual development package.

The output should explain both:

• Why this execution expresses the creative idea.
• What visual evidence the client can see before production.

Core Stance

• Write as a director speaking to a client, not as a production checklist.
• Be richer than a short style note: include emotion, image logic, camera grammar, performance, world, product, rhythm, sound, and visual proof.
• Use image generation as a treatment tool. For every major visual section, write a GPT-Image2-ready prompt and generate the corresponding image by default when image tools are available.
• Only skip image generation when the user explicitly asks for text/prompts only, or when the current environment cannot generate images.
• Keep generated visuals as references for approval, not final production frames unless the user asks.

Workflow

1. Confirm inputs:

• Approved creative idea or proposal route.
• Product truth, audience, channel, duration, visual references, actor policy, and AI/live-action mode.
• If these are missing and affect treatment direction, route back to $tvc-brief-strategist or $tvc-concept-developer.

1. Load visual development guidance when images are needed:

• Read references/visual-development.md.
• Use it to create visual deliverables and image-generation prompts.

1. Build the treatment:

• Director statement.
• Film promise and emotional arc.
• Image style and color system.
• Camera language and lens/movement grammar.
• Rhythm and edit feeling.
• Performance and persona design.
• Scene/world design.
• Product presentation.
• Art direction, lighting, texture.
• Music and sound.
• Reference and visual development plan.

1. Generate visual assets for each major section:

• Image style reference.
• Camera language reference.
• Performance/persona design, including character three-view when relevant.
• Scene setting image.
• Product presentation image.
• Lighting/color mood image.
• Key visual or hero frame.
• For each asset, output the GPT-Image2 prompt, generate the image, show it inline, and preserve the prompt used.

1. Hand off to:

• $tvc-ppm-producer for PPM expansion.
• $tvc-storyboard-writer for shot-by-shot structure.
• $tvc-ai-shot-prompter if visual prompts need per-shot conversion.

Output Format

## 导演阐述
 
## 影像风格
-文字阐述：
-视觉参考图 / GPT Image 2 Prompt：
 
## 镜头语言
-文字阐述：
-视觉参考图 / GPT Image 2 Prompt：
 
## 节奏与情绪曲线
-文字阐述：
-视觉参考图 / GPT Image 2 Prompt：
 
## 表演与人物形象设计
-文字阐述：
-人物三视图 / GPT Image 2 Prompt：
 
## 场景设定
-文字阐述：
-场景设定图 / GPT Image 2 Prompt：
 
## 产品呈现
-文字阐述：
-产品呈现图 / GPT Image 2 Prompt：
 
## 美术、光影与色彩
-文字阐述：
-光影色彩参考图 / GPT Image 2 Prompt：
 
## 音乐与声音
 
## 视觉开发清单
 
## 下游交接

Image Generation Rules

• Generate images by default for the visual sections when image tools are available.
• If the user asks for prompts only, do not generate images.
• If image tools are unavailable or blocked, provide GPT Image 2-ready prompts and explicitly state that image generation was not executed.
• For OpenAI image API implementation details, use $openai-docs when current API syntax matters.
• For normal in-chat image generation, use $imagegen behavior when available.
• Do not invent exact brand packaging details not present in the provided product reference.
• For product images, require reference-image fidelity and block logo/text hallucination.
• For character three-view, specify front, side, back views on one clean sheet, consistent wardrobe and face, neutral pose.
• For multiple treatment images, generate one asset per prompt rather than asking for a single collage unless the section specifically needs a beat-board or three-view sheet.

Quality Bar

• Every visual choice must serve the creative idea and audience psychology.
• Treatment writing must explain "why this visual language", not merely list adjectives.
• Each generated/reference image must answer a client question.
• Preserve continuity for later storyboard and AI prompting.
• Product presentation must protect shape, logo, package color, material, and scale.
• Character/persona design must match audience and avoid celebrity likeness unless explicitly approved.