流程梳理 / 质量评估 / 持续迭代PROCESS / JUDGEMENT / ITERATION

AI辅助设计与项目实践AI & Creative Practice

从需求分析到设计实施与功能验证。From requirements analysis to implementation and validation.

将AI应用于前期研究、视觉方案、素材制作与交互开发，通过项目案例展示工具参与、设计决策、效果修正与实际应用的完整过程。I work with AI across research, visual exploration, asset production and interactive development. These selected processes show where AI contributes, and where design judgement, practical skills and testing shape the outcome.

工作原则Working principles

01 / DESIGN JUDGEMENT

以设计判断把控输出质量Design judgement guides output quality

AI可以扩展制作能力，但生成结果仍存在不确定性与应用限制。我结合既有的设计与制作经验，判断输出是否符合项目要求；当结果未达到预期时，定位问题，并选择重新生成、手工调整或修改实现方案等处理方式。质量判断与修正能力，是我使用AI推进项目的基础。AI can extend production capabilities, but its outputs remain variable and subject to practical constraints. I draw on design and production experience to assess whether an output meets project requirements, identify gaps, and choose between regeneration, manual refinement or a revised implementation. Evaluation and corrective action underpin my use of AI.

02 / INPUT DEFINITION

以明确输入提高结果可控性Clear inputs support more controlled outcomes

使用AI前，我先梳理设计目标、视觉参考与应用要求，将需求转化为可执行的输入。在Merry Go Round中，手绘概念图比单纯文字描述提供了更明确的形态与色彩依据，帮助生成结果贴近预期；对于未达到要求的结果，再通过材质调整与后续制作完成修正。这是该项目中的实践观察。Before using AI, I clarify design goals, visual references and application requirements, then translate them into actionable inputs. In Merry Go Round, hand-drawn concept art provided more explicit shape and colour guidance than text alone. Outputs that fell short were refined through material adjustments and further production. This is an observation from that project.

查看AI数字影像Merry Go Round / View asset production case ↗

AI参与AI Contribution设计师参与Designer Contribution人机协同迭代Human–AI Iteration

工作流总览Workflow overview

A / AI辅助设计与开发AI-assisted Design & Development

01需求分析与方案策划Requirements Analysis & Design Planning↔02视觉方案与效果把控Visual Development & Quality Review↔03AI辅助素材制作AI-assisted Asset Production↔04交互开发与功能验证Interaction Development & Functional Testing

↶ 根据制作与测试反馈，回到前序环节调整方案Production and test findings feed back into earlier design decisions

B / AI功能整合与体验设计AI Integration & Experience Design

05 / EchoMemo · StepWise

从需求与交互设计阶段定义AI的作用，并在开发与测试中持续调整。此层展示AI作为体验功能的工作方式，可与上层设计开发流程协同。Define the role of AI during requirements and interaction design, then refine it through development and testing. This layer describes AI as an experience feature, supported by the design and development workflow above.

用户输入User inputAI处理与生成AI processing & generation交互反馈Interaction feedback

贯穿两层：设计判断 · 效果评估 · 迭代改进Across both layers: Design judgement · Evaluation · Iteration

明确要求 → 评估结果 → 定位差异 → 选择修正路径 → 再次验证Define requirements → Evaluate results → Identify gaps → Select corrective action → Validate again

两层代表AI辅助工作的方式与AI在体验中的作用，并非五个必须依次完成的步骤。点击节点可查看对应案例。The layers distinguish AI-assisted work from AI within an experience; they are not five mandatory sequential steps. Select a node to view its case.

01 /

需求分析与方案策划

Requirements Analysis & Design Planning

借助AI辅助资料梳理与概念讨论，结合案例分析与参考筛选，明确设计目标并形成可推进的方案。Use AI to support research synthesis and concept discussion, then define design goals through case analysis and reference selection.

AI辅助：问题讨论与资料梳理AI Contribution: discussion & research synthesis

设计师：参考筛选、内容分析与方案制定Designer Contribution: reference selection, interpretation & concept development

02 / 视觉方案与效果把控Visual Development & Quality Review

利用生成图像开展视觉预演，结合图像编辑、尺寸建模与场景制作，调整风格表现并检查方案的实施条件。Use generated imagery for visual studies, then refine the direction through image editing, dimensioned modelling and scene production.

AI辅助：效果图生成与视觉方向探索AI Contribution: visual studies & atmosphere exploration

设计师负责：风格设定、图像调整与制作落地Designer Contribution: visual direction, image editing & physical realisation

Case1 / 体验分镜与氛围预演Experience Storyboarding & Previsualisation

在项目前期，使用AI生成场景图像，并结合Illustrator与Photoshop调整不同阶段的氛围和风格，形成体验分镜，用于方案沟通与潜在设计问题的识别。Generate early scene studies with AI and refine atmosphere and style in Illustrator and Photoshop. Organise the results into an experience storyboard to communicate the proposal and identify potential design issues.

01 / 初始空间Initial environment

明确进入体验时的空间状态与视觉基调。Define the spatial conditions and visual tone at entry.

02 / 交互阶段Interaction sequence

对应用户动作，组织物件、声音与空间反馈。Map user actions to objects, sound and spatial responses.

03 / 氛围变化Atmosphere transition

比较色彩、清晰度与空间氛围的阶段变化。Review changes in colour, clarity and spatial atmosphere.

ROOM 完整体验分镜:用户所见、行为与感受/Full ROOM storyboard:what users see,do and feel

关键判断 / Key decision

将视觉变化与用户行为及预期情绪相对应，检查各阶段的体验衔接，为后续原型制作提供依据。Align visual changes with user actions and intended emotional responses. Review transitions between stages to inform subsequent prototyping.

相关项目 / Related project: ROOM ↗

Case 2 /空间方案与制作落地Spatial Design & Production

从视觉参考与AI效果预演出发，结合实际场地、材料与安装条件细化方案，将概念转化为尺寸模型、制作文档与实体装置。Develop visual references and AI studies into dimensioned models, production documents and a physical installation, informed by site, material and installation requirements.

视觉参考 → AI效果预演 → 尺寸模型与安装文档 → 实体搭建References → AI visual studies → Dimensioned models & installation documents → Physical installation

01 / 视觉参考Visual references

Visual Refrence

梳理半透明围合、材质与光线参考，明确空间视觉方向。Review translucent enclosures, materials and lighting to define the spatial direction.

02 / AI效果预演AI visual studies

AI辅助＋设计师调整 / AI-assisted + designer refinement

结合视觉参考生成效果图，经多轮调整形成概念预演。Generate visual studies from references and refine them through successive iterations.

03 / 尺寸模型与安装文档Dimensioned models & installation documents

结合场地条件进行尺寸建模，整理材料、连接方式与安装方案。Develop dimensioned models for the site and document materials, connections and installation options.

04 / 实体搭建Physical installation

将制作方案落实为实体装置，呈现空间与现场体验。Implement the production scheme as a physical installation and on-site experience.

关键判断 / Key decision

根据场地限制与制作条件调整方案，并明确材料与连接方式，使前期效果预演能够转化为实际可搭建的空间。Adapt the proposal to site and production constraints, specifying materials and connections to translate early visual studies into a buildable space.

相关项目 / Related project: ROOM ↗

Case 3/ 场景风格与制作策略Scene Direction & Production Planning

通过视觉参考与AI生成方案明确场景方向，结合项目要求与交付时间确定空间拼贴策略，并在Blender中完成素材整合与场景制作。Define the scene direction through references and AI-generated studies, select a spatial collage approach for the brief and timeframe, and integrate assets into a scene in Blender.

视觉参考 → 生成方案 → 拼贴策略 → Blender制作 → 最终效果References → Generated studies → Collage strategy → Blender production → Final visual

01 / 视觉参考Visual references

设计师：参考筛选 / Designer: reference selection

整理眼睛、地形与空间拼贴等视觉参考，形成场景方向。Select references for eyes, landscape and spatial collage to establish a scene direction.

02 / 生成方案对比Generated visual studies

AI辅助＋设计师评估 / AI-assisted + designer review

比较不同色彩、地形与氛围方案，明确后续制作的视觉依据。Compare colour, landscape and atmosphere variations to guide production.

03 / 空间拼贴策略Spatial collage strategy

设计师：制作策略 / Designer: production approach

结合项目主题、交付时间与既有制作经验，选择在三维空间中组织二维素材的表现方式。Choose to arrange 2D elements in a 3D environment based on the brief, delivery timeframe and production experience.

04 / Blender场景制作Scene production in Blender

AI素材＋设计师整合 / AI assets + designer integration

将AI生成的图像与模型素材整合进Blender，组织场景层次、空间关系与整体构图。Integrate AI-generated images and models in Blender, arranging scene layers, spatial relationships and composition.

05 / 最终场景效果Final scene visual

设计师：效果整合 / Designer: visual integration

汇总呈现场景构图、色彩和素材组合的最终效果。Present the final composition, colour and combined assets within the scene.

关键判断 / Key decision

在确定场景方向后，根据主题、交付时间与制作经验选择实现方式，将二维素材组织到三维空间中，使视觉方案能够进入实际制作。Choose an implementation method based on the theme, delivery timeframe and production experience, arranging 2D assets within a 3D space to move the visual proposal into production.

相关项目 / Related project: Dopamine Eyes

03 / AI辅助素材制作AI-assisted Asset Production

根据文字或概念图生成三维素材，在实际场景中检查效果，并通过材质、贴图与细节调整完成整合。Generate 3D assets from text or concept art, assess them in context, and refine materials, textures and details for integration.

AI辅助：三维模型生成AI Contribution: 3D model generation

设计师负责：概念图、材质调整与场景整合Designer Contribution: concept art, material refinement & scene integration

参考与输入 → 模型生成 → 场景检查 → 材质与贴图修正 → 最终应用Reference & input → Model generation → In-scene review → Material & texture refinement → Final integration

Case 01 / 输入方式与结果修正Case 01 / Input definition & output refinement

路径 A / 文字输入与后期修正Path A / Text input & subsequent refinement

01 / 文字输入与模型生成Text-to-3D generation

通过文字描述生成初始模型。Generate initial models from text descriptions.

02 / 场景应用与质量检查In-scene quality review

将模型放入场景，识别材质表现与预期的差异。Review the models in context and identify material mismatches.

03 / 着色器与材质修正Shader & Material Refinement

在Blender中调整着色器与材质，修正模型的表面表现。Adjust shaders and materials in Blender to refine the appearance of the models.

04 / 最终场景应用Final Scene Integration

将修正后的素材整合进场景，呈现最终效果。Integrate the refined assets into the scene and present the final result.

路径 B / 手绘概念图引导生成Path B / Generation guided by hand-drawn concept art

01 / 概念图定义Concept definition

以手绘概念图明确形态、配色与视觉风格。Define form, colour and visual style through hand-drawn concept art.

02 / 图像引导模型生成Image-to-3D generation

将概念图作为生成参考，检查输出与设计意图的一致性。Use the concept art as a generation reference and review alignment with the intended design.

03 / 结果评估与应用Output review & integration

本案例中，生成结果的材质与风格符合预期，进入最终场景应用。In this case, the generated materials and style met expectations and were used in the final scene.

相关项目 / Related project: Merry Go Round

案例 02 / 面向独立式VR的模型与贴图优化
Case 02 / Mesh & texture optimisation for standalone VR

从初始模型到场景应用From initial model to scene integration

01 / 初始模型生成Initial model generation

通过文字输入生成三维模型，作为后续制作起点。Generate a 3D model from text as the starting point for production.

02 / 场景检查与模型减面In-scene Review & Polygon Reduction

在场景中检查模型表现，并进行减面处理，降低几何复杂度，为独立式VR设备的运行环境适配素材。Review the model in context and reduce its polygon count to lower geometric complexity for use on standalone VR devices.

03 / Substance贴图制作Texture Production in Substance

在Substance中制作和调整贴图，完善模型的材质与表面细节。Create and refine textures in Substance to develop the materials and surface details.

04 / 最终场景应用Final Scene Integration

将优化后的模型与贴图整合进场景，检查材质表现与整体视觉效果。Integrate the optimised model and textures into the scene and review the materials and overall visual result.

相关项目 / Related project: The Line She Made

04 / Vibe Coding 交互开发与功能验证Vibe Coding - Interaction Development & Functional Testing

借助AI辅助代码实现与问题排查，通过运行测试验证交互功能，并记录配置步骤、迭代结果及待处理问题。Use AI to assist implementation and troubleshooting, validate interactions through runtime tests, and document configuration, iterations and unresolved issues.

AI辅助：代码建议与问题排查AI Contribution: coding assistance & troubleshooting suggestions

设计师负责：实现方案、运行测试与文档整理Designer Contribution: implementation decisions, runtime testing & documentation

多人交互开发、调试与测试记录

Multiplayer development, debugging and testing log · Select to view at full size

05 / AI功能整合与生成质量优化AI Integration & Generative Output Refinement

EchoMemo

将语音记忆转化为可交互的MR视觉内容。通过ConvAI与OpenAI接口接入Unity，串联对话引导、记忆信息提取、图像生成与空间呈现，并针对图像生成环节开展提示词迭代与场景内评估。Translate spoken memories into interactive visual content in MR. Connect ConvAI and OpenAI APIs to Unity to support guided dialogue, memory extraction, image generation and spatial presentation, with prompt iteration and in-scene evaluation focused on the image-generation stage.

Unity · ConvAI · OpenAI APIs · Prompt Iteration · Visual Evaluation

AI参与：对话引导、信息提取与图像生成AI contribution: guided dialogue, information extraction and image generation

设计与开发：交互流程、接口整合、提示词调试与视觉评估Design & development: interaction flow, API integration, prompt refinement and visual evaluation

01 / 体验流程与AI接口整合Experience Flow & AI Integration

流程图展示用户输入与AI处理之间的衔接：用户讲述记忆，AI引导并提取上下文；信息不足时继续追问，信息完整后生成视觉记忆，再进入存储与回放环节。The flow maps user input to AI processing: users narrate a memory, the assistant guides the conversation and extracts context, follow-up questions address missing information, and visual generation leads into storage and replay.

EchoMemo体验流程：用户操作与AI处理EchoMemo experience flow: user actions and AI processing

02 / 图像生成提示词迭代与效果评估Image-generation Prompt Iteration & Evaluation

围绕流程中的图像生成环节，我将视觉目标转化为提示词约束，生成多个版本并放入Unity场景进行比较。通过调整风格描述、模糊程度与人物具体性，提高输出的视觉一致性与可控性。For the image-generation stage, I translated visual goals into prompt constraints, generated variations and compared them within Unity. I refined style descriptions, blur and character specificity to improve visual consistency and controllability.

风格约束Style constraints

将“复古”拆解为明确的视觉属性，减少单一风格词带来的歧义。Define specific visual attributes rather than relying on the word “vintage” alone.

清晰度与氛围Clarity & atmosphere

比较不同模糊程度，平衡怀旧氛围、内容辨识度与MR可读性。Compare levels of blur to balance nostalgia, recognisability and readability in MR.

人物具体性Character specificity

比较具体面孔与物件、手势和局部人物的表达，筛选适合个人记忆投射的画面。Compare specific faces with objects, gestures and partial figures to select imagery suited to personal memory projection.

提示词调试与视觉评估记录

Prompt refinement and visual evaluation record

补充案例Additional projectAI+XR 智能维修助手 StepWise

项目复盘 / 工作方法迭代REFLECTION / AN ONGOING PRACTICE

明确目标 · 评估质量 · 迭代改进Define. Evaluate. Refine.

根据项目目标明确设计要求，结合使用场景评估生成结果，并选择合适的修正方法，推进设计实施与项目交付。Tools and production methods continue to change. Across these projects, I focus on defining the intended experience, assessing the output and choosing how to move the work forward.

回看2024年项目 ↗Return to the 2024 reflection ↗