[{"data":1,"prerenderedAt":22},["ShallowReactive",2],{"blog-scientific-figures-gpt-image2-prompt-guide":3},{"id":4,"title":5,"title_fr":6,"slug":7,"summary":8,"summary_fr":6,"cover_image":9,"content":10,"content_fr":6,"tags":11,"status":19,"sort_order":4,"created_at":20,"updated_at":21},22,"GPT Image 2 for Scientific Figures: Prompt Design & Editable Workflow Guide",null,"scientific-figures-gpt-image2-prompt-guide","A complete guide to creating publication-quality scientific figures with GPT Image 2. Covers prompt templates, figure type matching, visual hierarchy, SVG conversion workflow, and design principles aligned with Nature, Science, and IEEE standards.","\u002Fdata\u002Fimages\u002Fscientific-figure-workflow.jpg","\u003Ch2>Introduction\u003C\u002Fh2>\n\u003Cp>Stop simply copying your paper abstract into AI image generators — that is where most people go wrong when creating scientific figures with AI. For researchers, a scientific figure is not merely an illustration. It is an essential part of the storytelling process of a paper.\u003C\u002Fp>\n\u003Cp>A well-designed figure performs a form of \u003Cstrong>information compression\u003C\u002Fstrong>: within a limited visual space, it helps readers quickly understand: what problem the study addresses; what methodology is proposed; where the key innovation lies; why the proposed approach works.\u003C\u002Fp>\n\u003Cp>The core workflow should be: \u003Cstrong>Paper Understanding → Figure Type Selection → Information Abstraction → Visual Encoding\u003C\u002Fstrong>\u003C\u002Fp>\n\n\u003Cdiv style=\"text-align:center;margin:30px 0\">\u003Cimg src=\"\u002Fdata\u002Fimages\u002Fscientific-figure-workflow.jpg\" alt=\"Scientific Figure Generation Workflow\" style=\"max-width:100%;border-radius:12px;box-shadow:0 4px 20px rgba(0,0,0,0.1)\"\u002F>\u003Cp style=\"color:#888;font-size:13px;margin-top:8px\">Figure 1: AI-Powered Scientific Figure Generation Pipeline\u003C\u002Fp>\u003C\u002Fdiv>\n\n\u003Ch2>Our Approach: Making AI Understand Research Logic\u003C\u002Fh2>\n\u003Cp>A better approach is to explicitly instruct the model to analyze: the research problem; the technical pipeline; the core contributions; the experimental validation logic; the relationship between different components. The process: \u003Cstrong>Paper Content → Research Logic Analysis → Visual Design → Scientific Figure Generation\u003C\u002Fstrong>\u003C\u002Fp>\n\n\u003Ch2>Scientific Figure Types Should Match Research Paradigms\u003C\u002Fh2>\n\u003Ctable>\n\u003Ctr>\u003Cth>Research Paradigm\u003C\u002Fth>\u003Cth>Figure Type\u003C\u002Fth>\u003Cth>Visual Focus\u003C\u002Fth>\u003C\u002Ftr>\n\u003Ctr>\u003Ctd>Theoretical proposal\u003C\u002Ftd>\u003Ctd>Conceptual framework\u003C\u002Ftd>\u003Ctd>New concepts, hypotheses\u003C\u002Ftd>\u003C\u002Ftr>\n\u003Ctr>\u003Ctd>Algorithm design\u003C\u002Ftd>\u003Ctd>Method overview, pipeline\u003C\u002Ftd>\u003Ctd>Problem, solution strategy\u003C\u002Ftd>\u003C\u002Ftr>\n\u003Ctr>\u003Ctd>Deep learning architecture\u003C\u002Ftd>\u003Ctd>Network architecture\u003C\u002Ftd>\u003Ctd>Components, info flow\u003C\u002Ftd>\u003C\u002Ftr>\n\u003Ctr>\u003Ctd>Component improvement\u003C\u002Ftd>\u003Ctd>Module zoom-in, comparison\u003C\u002Ftd>\u003Ctd>Structural changes\u003C\u002Ftd>\u003C\u002Ftr>\n\u003Ctr>\u003Ctd>Reasoning mechanism\u003C\u002Ftd>\u003Ctd>Information flow, causal graph\u003C\u002Ftd>\u003Ctd>Internal processing\u003C\u002Ftd>\u003C\u002Ftr>\n\u003Ctr>\u003Ctd>Dataset engineering\u003C\u002Ftd>\u003Ctd>Data lifecycle pipeline\u003C\u002Ftd>\u003Ctd>Collection, processing\u003C\u002Ftd>\u003C\u002Ftr>\n\u003Ctr>\u003Ctd>Benchmark\u002Fevaluation\u003C\u002Ftd>\u003Ctd>Evaluation framework\u003C\u002Ftd>\u003Ctd>Tasks, metrics, protocols\u003C\u002Ftd>\u003C\u002Ftr>\n\u003Ctr>\u003Ctd>LLMs\u002Fmultimodal\u003C\u002Ftd>\u003Ctd>Multimodal framework\u003C\u002Ftd>\u003Ctd>Data fusion, knowledge\u003C\u002Ftd>\u003C\u002Ftr>\n\u003Ctr>\u003Ctd>Agents\u002Fembodied AI\u003C\u002Ftd>\u003Ctd>Closed-loop workflow\u003C\u002Ftd>\u003Ctd>Perception, action, feedback\u003C\u002Ftd>\u003C\u002Ftr>\n\u003Ctr>\u003Ctd>Cross-disciplinary AI\u003C\u002Ftd>\u003Ctd>Domain-AI framework\u003C\u002Ftd>\u003Ctd>Domain knowledge, validation\u003C\u002Ftd>\u003C\u002Ftr>\n\u003C\u002Ftable>\n\n\u003Cdiv style=\"text-align:center;margin:30px 0\">\u003Cimg src=\"\u002Fdata\u002Fimages\u002Fscientific-figure-types.jpg\" alt=\"Scientific Figure Types Table\" style=\"max-width:100%;border-radius:12px;box-shadow:0 4px 20px rgba(0,0,0,0.1)\"\u002F>\u003Cp style=\"color:#888;font-size:13px;margin-top:8px\">Figure 2: Research Paradigm to Figure Type Mapping\u003C\u002Fp>\u003C\u002Fdiv>\n\n\u003Ch2>Visual Style: Nature, Science, IEEE Standards\u003C\u002Fh2>\n\u003Cp>Clean 2D vector style; White background; Minimal noise; Limited color palette; Colors only for functional grouping. Output should resemble Method Overview figures, Framework diagrams — not PowerPoint slides or marketing graphics.\u003C\u002Fp>\n\n\u003Ch2>Visual Hierarchy Matters\u003C\u002Fh2>\n\u003Cp>\u003Cstrong>Highest priority:\u003C\u002Fstrong> Core innovation, proposed method. \u003Cstrong>Second:\u003C\u002Fstrong> Overall workflow, data relationships. \u003Cstrong>Supporting:\u003C\u002Fstrong> Input data, experiments, results.\u003C\u002Fp>\n\n\u003Ch2>Complete Prompt Template\u003C\u002Fh2>\n\u003Cpre>\u003Ccode>You are a professional scientific visualization designer with expertise in CS, AI, engineering systems, and data science. Design publication-quality figures following Nature\u002FScience\u002FIEEE standards.\n\nAnalyze: research background; scientific question; research paradigm; technical methodology; data flow; experimental validation; main contributions.\n\nThe figure should communicate: What problem is addressed; What method is proposed; Why the method works; What value it provides.\n\nSelect figure structure based on research type. Apply strict visual hierarchy. Use clean 2D vector style, white background, limited palette. Colors only for functional grouping.\u003C\u002Fcode>\u003C\u002Fpre>\n\n\u003Ch2>Making AI Figures Editable (SVG Workflow)\u003C\u002Fh2>\n\u003Cp>Simply ask: \u003Cem>\"Convert this scientific figure into SVG code. Keep every module, label, arrow, and icon as an independently editable element.\"\u003C\u002Fem> Then refine in Adobe Illustrator, Figma, or Inkscape.\u003C\u002Fp>\n\n\u003Ch2>Recommended Paper Submission Workflow\u003C\u002Fh2>\n\u003Cp>\u003Cstrong>GPT Image 2 → Initial Design → SVG Conversion → Illustrator\u002FFigma Refinement → Publication-Ready Figure\u003C\u002Fstrong>\u003C\u002Fp>\n",[12,13,14,15,16,17,18],"GPT Image 2"," scientific figures"," prompt engineering"," AI visualization"," academic publishing"," figure design"," SVG workflow",1,"2026-07-01T08:52:49.000Z","2026-07-01T09:47:29.000Z",1782899343253]