Auto-sync: 2026-04-20 07:08
This commit is contained in:
55
wiki/sources/design-inclusive-visuals-specialist.md
Normal file
55
wiki/sources/design-inclusive-visuals-specialist.md
Normal file
@@ -0,0 +1,55 @@
|
||||
---
|
||||
title: "Inclusive Visuals Specialist"
|
||||
type: source
|
||||
tags: [The Agency, AI Agent, Design]
|
||||
date: 2026-04-20
|
||||
---
|
||||
|
||||
## Source File
|
||||
- [[raw/Agent/agency-agents/design/design-inclusive-visuals-specialist.md]]
|
||||
|
||||
## Summary
|
||||
- 核心主题：AI 图像与视频生成中的包容性视觉设计专家智能体
|
||||
- 问题域：解决基础图像/视频模型（Midjourney、Sora、Runway、DALL-E）中的系统性刻板印象和偏见问题
|
||||
- 方法/机制：通过精确的提示词工程、负向约束库、物理现实定义等技术手段，生成具有文化准确性、尊严和真实感的多元人像
|
||||
- 结论/价值：确保 AI 生成媒体以尊严、主体性和真实情境现实主义描绘各类人群，对抗默认偏见和 AI 幻觉
|
||||
|
||||
## Key Claims
|
||||
- 身份不应被视为简单的描述符输入，而是需要专业技术准确表达的领域
|
||||
- 必须明确要求不同群体中的面部结构、年龄和体型各异，防止生成"克隆面孔"
|
||||
- 必须显式负向提示任何文本、标志或生成标牌，防止 AI 发明令人反感或无意义的字符
|
||||
- 在视频生成中必须明确定义服装、头发和助行辅助工具的物理特性
|
||||
|
||||
## Key Quotes
|
||||
> "The current prompt will likely trigger the model's 'exoticism' bias. I am injecting technical constraints to ensure the lighting and geographical architecture reflect authentic lived reality." — 关键技术短语
|
||||
|
||||
> "You review AI output not just for technical fidelity, but for sociological accuracy." — 核心工作标准
|
||||
|
||||
## Key Concepts
|
||||
- [[InclusiveVisualsSpecialist]]：专注于真实人类representation的严格提示词工程师，对抗基础图像和视频模型中嵌入的系统性刻板印象
|
||||
- [[CloneFaces]]：AI 在生成多元群体时生成多个相同面孔的问题，需要通过显式约束防止
|
||||
- [[GibberishText]]：AI 在尝试非英语脚本或文化符号时发明无意义或冒犯性字符的问题
|
||||
- [[PhysicalRealityConstraints]]：在视频生成中明确定义服装、头发和助行辅助工具物理特性的技术
|
||||
- [[NegativePromptLibrary]]：针对图像和视频平台的显式负向提示库，用于阻止"AI 怪异感"
|
||||
|
||||
## Key Entities
|
||||
- [[TheAgency]]：开源 AI 智能体集合项目，汇集各类专业化 AI Agent
|
||||
- [[InclusiveVisualsSpecialist]]：The Agency 项目中的包容性视觉设计专家智能体
|
||||
- Midjourney：AI 图像生成模型
|
||||
- Sora：OpenAI 视频生成模型
|
||||
- Runway：AI 视频生成平台
|
||||
- DALL-E：OpenAI 图像生成模型
|
||||
|
||||
## Connections
|
||||
- [[InclusiveVisualsSpecialist]] ← belongs_to ← [[TheAgency]]
|
||||
- [[InclusiveVisualsSpecialist]] ← depends_on ← [[ImagePromptEngineer]]
|
||||
- [[CloneFaces]] ← counters ← [[Midjourney]]
|
||||
- [[CloneFaces]] ← counters ← [[Sora]]
|
||||
- [[GibberishText]] ← counters ← [[DALL-E]]
|
||||
- [[PhysicalRealityConstraints]] ← requires ← [[NegativePromptLibrary]]
|
||||
|
||||
## Contradictions
|
||||
- 与"批量生成多样性的过度纠正"冲突：
|
||||
- 冲突点：AI 尝试"过于多样化"时会生成符号化、不真实的构图
|
||||
- 当前观点：需要精确的技术约束来平衡真实性和多样性
|
||||
- 对方观点：简单地堆叠身份描述符即可实现多样性
|
||||
Reference in New Issue
Block a user