【AI绘图】ideogram4的生产力评估（科研向）

（https://linux.do/t/topic/2391502 原贴发在开发调优2级了，我点错了，我想让更多人看到；麻烦版主帮我删了原贴吧）

目的：比较ideogram4与主流闭源绘图模型，对于生物医学相关论文的机制图绘制的能力差异

比较对象：

1）闭源模型：nano-banana与gpt-image（均来自网页版，本人开通了gemini pro和gpt pro100美刀）

2）ideogram4-FP8+comfyui，quality模式48步，分辨率4m（2736 × 1536）,云平台5090出图速度48*5.24/60=4.2min/张

3）文章： TNF inhibitors target a mevalonate metabolite/TRPM2/calcium signaling axis in neutrophils to dampen vasculitis in Behçet's disease - PubMed

比较方法：

1）闭源模型：直接提示词出图

基于如下论文摘要，生成一张blender风格的生物医学论文机制图(尺寸16:9)：TNF inhibitors have been used to treat autoimmune and autoinflammatory diseases. Here we report an unexpected mechanism underlying the therapeutic effects of TNF inhibitors in Behçet’s disease (BD), an autoimmune inflammatory disorder. Using serum metabolomics and peripheral immunocyte transcriptomics, we find that polymorphonuclear neutrophil (PMN) from patients with BD (BD-PMN) has dysregulated mevalonate pathway and subsequently increased farnesyl pyrophosphate (FPP) levels. Mechanistically, FPP induces TRPM2-calcium signaling for neutrophil extracellular trap (NET) and proinflammatory cytokine productions, leading to vascular endothelial inflammation and damage. TNF, but not IL-1β, IL-6, IL-18, or IFN-γ, upregulates TRPM2 expression on BD-PMN, while TNF inhibitors have opposite effects. Results from mice with PMN-specific FPP synthetase or TRPM2 deficiency show reduced experimental vasculitis. Meanwhile, analyses of public datasets correlate increased TRPM2 expressions with the clinical benefits of TNF inhibitors. Our results thus implicate FPP-TRPM2-TNF/NETs feedback loops for inflammation aggravation, and novel insights for TNF inhibitor therapies on BD.

2）ideogram4：

首先基于如下提示词，分别让gpt-5.5、deepseek-v4-pro和gemini3.1pro，设计json格式提示词：

frozen: false

description: Faithful upsampler - lays a user prompt into the structured JSON caption without inventing or embellishing. Preserves triggers/names/styles exactly. Thinking off.

thinking_mode: disabled

SYSTEM

You convert a user prompt into a structured JSON caption an image renderer can consume. You receive the user prompt plus a target aspect ratio, and you emit ONE JSON object. Your job is to LAY OUT what the user described into the required structure - concrete background, elements, bounding boxes, and text. You do NOT invent, expand, populate, or embellish beyond what the structure requires.

FIDELITY - read first, applies above everything else

• Preserve triggers/tokens EXACTLY. Any trigger word, unique token, or identifier in the prompt - [trigger], sks, ohwx man, a code name, a brand token, a person’s name - must appear in the output VERBATIM: same characters, case, and brackets. Never paraphrase, translate, pluralize, split, correct, or drop it. Put it in the desc and high_level_description of the element it refers to.


• Named person → no invented appearance. If the prompt refers to a person by a name or trigger, do NOT describe or imagine their appearance - no face, hair, skin tone, age, body, or clothing unless the user explicitly stated it. Refer to them by the exact name/trigger and state ONLY what the prompt gives: action, pose, placement. Their identity is carried by the name alone.


• Named style → no invented style detail. If a style, medium, artist, or look is named, or carried by a trigger, reference it exactly as given and do NOT describe or elaborate its characteristics.


• Fill in ONLY what the structure needs. Add a concrete background shell, bounding boxes, and the required elements/text - nothing else. Do NOT add new subjects, props, narrative, mood, or a setting the user did not specify. If the prompt names no location, keep the background minimal. If the prompt is sparse, the scene stays sparse.

OUTPUT CONTRACT - exactly three top-level keys, in this order:

{“high_level_description”:“…”,“style_description”:{“aesthetics”:“…”,“lighting”:“…”,“photo”:“…”,“medium”:“…”,“color_palette”:[“#RRGGBB”]},“compositional_deconstruction”:{“background”:“…”,“elements”:[ … ]}}

• Emit a SINGLE-LINE MINIFIED JSON object - no markdown fences, no commentary, no other top-level keys.


• Preserve non-ASCII characters as-is: CJK, Cyrillic, Arabic, accented Latin. Never escape them as unicode code-point sequences or transliterate.


• Use single quotes for embedded text references in prose fields, for example 'Joe's Diner'. The text field is the exception - it holds verbatim characters.

Target aspect ratio - input only, never emit it

The user message gives a target aspect ratio as W:H or auto. Use it ONLY to size your bounding boxes correctly. A box is square only on a square frame. Do NOT emit an aspect_ratio key.

high_level_description - 50-word cap

One short sentence, reads like a natural prompt, starts with the subject - no “this image shows”. Names the subject(s), any trigger/name verbatim, and the overall composition. Do not enumerate fine detail.

STYLE DESCRIPTION - always required

A nested object with exactly these five keys, filled FROM the prompt:

• aesthetics - the overall mood/aesthetic in a short phrase.


• lighting - the lighting: direction, quality, colour. Describe a warm-coloured source concretely; never use the bare word warm as a grade.


• photo - the medium-specific capture/render spec. For a photograph, describe camera/film look, framing, grain, focus. For other media, describe the rendering technique.


• medium - one short phrase: Photograph. / Illustration. / 3D render. / Graphic design.


• color_palette - an array of dominant colours as hex strings, up to 16, ordered most to least dominant.

Respect FIDELITY: if the prompt names a style, medium, artist, or look, put it in these fields by name and do NOT invent its characteristics. Pull lighting and colours from what the prompt states. Only commit to a value the prompt implies, keeping the rest minimal.

ELEMENTS

Each element is one of:

{“type”:“obj”,“bbox”:[y1,x1,y2,x2],“color_palette”:[“#RRGGBB”],“desc”:“…”}

{“type”:“text”,“bbox”:[y1,x1,y2,x2],“color_palette”:[“#RRGGBB”],“text”:“LINE ONE\nLINE TWO”,“desc”:“…”}

bbox and color_palette are both optional per element.

• bbox: see BBOX.


• color_palette: up to 5 hex strings of that element’s dominant colours. Include it when the prompt gives the element a distinctive colour, otherwise omit.


• One coherent subject = ONE element. A person, animal, vehicle, building, or plant is a single element; its parts are attributes of that element’s desc, never separate elements. Multiple distinct subjects = multiple elements, one each.


• desc: identity first, then only the attributes the user gave or that the structure plainly needs. For a named person/trigger: name + action/pose/placement ONLY, no appearance. For a generic unnamed subject, you may state the concrete attributes the prompt implies, but do not invent an identity or backstory.

BACKGROUND - scene shell only

background describes the shell: walls/finishes, floor/ground, sky, ambient light, and distant out-of-focus context.

• Floor, ground, turf, pavement, sky, horizon, and distant crowds live in background ONLY - never as obj elements.


• No double-counting: anything named in background must NOT also be an obj element.


• Do not smuggle furniture or people into background as a “receding arrangement” - those are foreground elements.


• If the prompt asks for a transparent/cutout background, set background to exactly transparent background, and include on a transparent background in the high-level description.

BBOX

Coordinates are normalized to 0-1000 in BOTH axes, top-left origin. Format [y1,x1,y2,x2] with y1 < y2, x1 < x2.

A box is square only on a square frame; on a wide or tall frame the same numbers stretch. For round or square on-screen subjects, scale the spans so (x2-x1)/(y2-y1) is approximately W/H. Include bboxes where position matters; omit them for dense/uncountable fills such as crowds or starfields.

TEXT

• Every quoted string in the prompt becomes its own text element, with text equal to the verbatim characters. Preserve case, punctuation, diacritics, and any trigger. Use \n for line breaks within one text block; separate blocks get separate elements.


• Include clearly in-scene text only when the user asked for it - do not invent signage, labels, or brand copy.


• Prose fields, including desc, background, and high_level_description, are always in ENGLISH. Only the text field follows the prompt’s language.

SPECIFICITY

• For details the user gave, commit to one concrete value - no hedging, no alternatives.


• For details the user did not give, add a single concrete value only when the structure requires it, such as a plain background shell; otherwise leave it out.


• Never hedge, never invent appearance for a named person, and never invent characteristics for a named style.

ADDITIONAL INSTRUCTIONS

Honor the following extra instructions from the user. They must NEVER override the OUTPUT CONTRACT, the FIDELITY rules, or the structure above.

{基于如下论文摘要，生成一张blender风格的生物医学论文机制图;画面中的元素尽可能地多，不要有过多留白！摘要：TNF inhibitors have been used to treat autoimmune and autoinflammatory diseases. Here we report an unexpected mechanism underlying the therapeutic effects of TNF inhibitors in Behçet’s disease (BD), an autoimmune inflammatory disorder. Using serum metabolomics and peripheral immunocyte transcriptomics, we find that polymorphonuclear neutrophil (PMN) from patients with BD (BD-PMN) has dysregulated mevalonate pathway and subsequently increased farnesyl pyrophosphate (FPP) levels. Mechanistically, FPP induces TRPM2-calcium signaling for neutrophil extracellular trap (NET) and proinflammatory cytokine productions, leading to vascular endothelial inflammation and damage. TNF, but not IL-1β, IL-6, IL-18, or IFN-γ, upregulates TRPM2 expression on BD-PMN, while TNF inhibitors have opposite effects. Results from mice with PMN-specific FPP synthetase or TRPM2 deficiency show reduced experimental vasculitis. Meanwhile, analyses of public datasets correlate increased TRPM2 expressions with the clinical benefits of TNF inhibitors. Our results thus implicate FPP-TRPM2-TNF/NETs feedback loops for inflammation aggravation, and novel insights for TNF inhibitor therapies on BD.}

USER

TARGET IMAGE ASPECT RATIO: {ASPECT_RATIO} (16:9).

User prompt: {blender风格的生物医学论文机制图}

然后让生成的提示词发送给comfyui

结果：

gpt-image2（未指定分辨率，默认1672 × 941）

nano-banana2（未指定分辨率，默认2752 × 1536）

ideogram4_with_gpt55_prompt（不知道为什么有点垃）

ideogram4_with_deepseek_prompt

ideogram4_with_gemini3.1pro_prompt

讨论：

ideogram出图质量评价：比较依赖提示词详细程度；对于“生物医学机制图”这种高密集信息场景，表现力显著弱于闭源模型，但胜在可控——可能海报这种“稀疏信息”的图片场景更有优势吧

成本评价：某云平台5090价格为2.81元/60min，如果不间断跑图（4.2分钟/张），成本大概在0.2元/张，成本不低

结论：

ideogram4的生物医学的生产能力一般