Click Reader · Method

Taste as Method

Our images aren't prompted — they're taught, by a roomful of artists.

Three reusable assets behind every image

Click Reader doesn't generate images from a single prompt. It composes them out of three reusable assets — visual, knowledge, and mechanism — each callable by the pipeline.

01 · Visual asset

405

Art reference library

Hundreds of curated photographs, paintings, and film stills. Each tagged with an 11-dimension visual signature (light, composition, medium, era, culture, mood…) and recombined into the picture this book deserves.

02 · Knowledge asset

691

Photography / painting RAG

We break visual language into entries an LLM can search: light direction, compositional rhythm, lens grammar, color psychology. Agents reach for these when writing prompts.

03 · Mechanism asset

13

Workflow + agent design

Seven pipeline steps + five agent decisions. Assets aren't piled up — they're scheduled by a system where every step is observable, replayable, tunable.

Still growing — tell us what is missing.

Art reference library

Wall arrangement, density first. Every image here makes an aesthetic claim before Click Reader translates books into pictures.

莫奈风格插画
Bazille and Camille (Study for "Déjeuner sur l'Herbe") - Claude Monet

温斯洛·霍默风格插画
Hauling in the Nets - Winslow Homer

莫奈风格插画
Woman with a Parasol - Madame Monet and Her Son - Claude Monet

莫奈风格插画
Waterloo Bridge, London, at Sunset - Claude Monet

温斯洛·霍默风格插画
Sketch for "Hound and Hunter" - Winslow Homer

莫奈风格插画
Waterloo Bridge, London, at Dusk - Claude Monet

温斯洛·霍默风格插画
View of Santiago de Cuba - Winslow Homer

莫奈风格插画
Rouen Cathedral, West Façade, Sunlight - Claude Monet

温斯洛·霍默风格插画
Native hut at Nassau - Winslow Homer

莫奈风格插画
Rouen Cathedral, West Façade - Claude Monet

达芬奇风格插画
Ginevra de' Benci [obverse] - Leonardo da Vinci

修拉风格插画
Study after "The Models" - Georges Seurat

马奈风格插画
The Tragic Actor (Rouvière as Hamlet) - Édouard Manet

柯罗风格插画
Italian Peasant Boy - Jean-Baptiste-Camille Corot

克林姆特风格插画
Curled up Girl on Bed - Gustav Klimt

德拉克洛瓦风格插画
Portrait of a Child - Eugène Delacroix

雷诺阿风格插画
Claude Renoir ("Coco") - Pierre-Auguste Renoir

维米尔风格插画
Woman Holding a Balance - Johannes Vermeer

高更风格插画
Three Tahitians (Study for "La soeur de charite") - Paul Gauguin

鲁本斯风格插画
The Capitoline Eagle - Peter Paul Rubens

毕加索风格插画
L'atelier de Cannes (The Studio in Cannes) - Pablo Picasso

康斯太勃尔风格插画
Salisbury Cathedral from Lower Marsh Close - John Constable

波提切利风格插画
The Virgin Adoring the Child - Sandro Botticelli

弗里德里希风格插画
Untitled from Eight Variations for Galerie Heiner Friedrich (Acht Variationen für die Galerie Heiner Friedrich) - Caspar David Friedrich

Hover an image to see its visual signature (light, composition, medium…). Filter groups derive from style_configs.display_name.

Photography / painting RAG

Visual language broken into entries an LLM can retrieve. Agents don't lean on general knowledge when writing prompts — they pull from here. So every choice has a why.

Light

'Photographing' your work
 > **AI 视觉分析**: 雕塑侧脸特写，冷蓝光打亮轮廓，暗背景强化立体感。风格写实而富有戏剧性，色彩对比强…
Crafty arts paper & textiles
 > **AI 视觉分析**: 该图像为一件抽象雕塑，主体呈球形，以黑色为主调，表面嵌有银色反光碎片，形成破碎感。风格前卫、工业感强，构图居中对称，色彩对比强烈，突显材…
Crafty arts ceramics & glass
 > **AI 视觉分析**: 雕塑人像侧脸，面部有金色裂痕，金属质感，暗调背景。风格具象而富有象征性，色彩以灰金为主，构图紧凑，突出裂痕与静谧表情，传递修复或伤痕美学…
Unexpected Applications
> **AI 视觉分析**: 玻璃幕墙透射彩虹色光，营造梦幻光影效果。热带植物与石景环绕水池，现代建筑与自然融合。色彩饱和绚丽，构图对称而富有层次，反射与透射交织，呈现超现实艺术风格。 ](https://midlibr…
Lighting prompts, artificial indoor
Studio lighting , professional lighting. studio portrait, well-lit, etc (source) Defined 'real' light source (…
Basic Prompting – The Essentials › Color & Lighting
We have full control over both color and lighting in our images. This can include the **color** of particular…

Composition

Camera angles proximity
 > **AI 视觉分析**: 一位老年男性专注下国际象棋，特写构图突出其沉思神情。浅蓝衬衫配暖棕棋子，背景素净…
Modern art movements II
 > **AI 视觉分析**: 画面描绘一只戴墨镜的白猫悠闲坐在木椅上，旁有奇异机械装置。风格超现实，色彩沉郁偏暖，构图居中对称，营造荒诞静谧氛围，具象征意味。 Fut…
Camera angles position
Tilted frame , dutch angle, skewed shot  > **AI 视觉分析**:…
Camera settings + lenses
 > **AI 视觉分析**: 一只企鹅跃出水面，姿态动感十足。画面以黑白为主调，点缀橙色喙与脚蹼，水花四溅营造张力。构图聚焦主体，背景虚化，风格写实而富有冲击力，凸显…
Edits uncropping, context jumping
We can also put our source image in a different context: in a frame on a wall, as a reflection in someone's ey…
Places and spaces
Public realm , streets, squares, plazas, cityscapes  > *…

Lens

Camera settings + lenses
 > **AI 视觉分析**: 卡通风格的灰色小生物立于绿叶与暖光中，大眼睛圆润可爱，表情惊讶。背景虚化暖黄光斑，前景鲜绿植物衬托主体，色彩柔和温馨，构图居中聚焦，营造…
Camera settings + lenses
Fish-eye lens: distorts the scene, vv. wide angle, the centre 'bulges' Motion blur (Source) <!-- IMG_META: pag…
Camera settings + lenses
 > **AI 视觉分析**: 一只企鹅跃出水面，姿态动感十足。画面以黑白为主调，点缀橙色喙与脚蹼，水花四溅营造张力。构图聚焦主体，背景虚化，风格写实而富有冲击力，凸显…
Camera angles position
Tilted frame , dutch angle, skewed shot  > **AI 视觉分析**:…
Camera angles proximity
Film still, extreme wide shot of an elephant alone on the savannah, extreme long shot <!-- IMG_META: page=19 -…
Camera angles proximity
DALL·E interprets these pretty loosely, and often provides candidate from the 'neighbour' framing, e.g: a clos…

Color

Emotional prompt words
 > **AI 视觉分析**: 油画风格肖像，描绘一男孩身着鲜红外套，侧坐椅上低头凝思。色彩以暖红为主调，背景柔和朦胧，笔触粗犷奔放，构图紧凑聚焦人物神态，传递静谧沉思…
Blending Styles
](https://midlibrary.io/focus/liu-ye) Liu Ye's collaborative painting with Takashi Murakami --v 6 Liu Ye maste…
Edvard Munch › Pro Tips
Use color choices that match the desired emotion: red and yellow for passion, blues and greens for melancholy.…
Looks, vibes, -punks, -waves
 > **AI 视觉分析**: 画面展现一座被苔藓覆盖的奇幻木屋，隐于浓雾森林中。风格奇幻写实，色调以冷绿灰为主，营造神秘幽静氛围。构图以倾斜树干为前景引导视线，低角度…
Liu Ye was born in 1964 in Beijing during the Cult
o.%202%20%281998%29.jpg) > **AI 视觉分析**: 超现实主义画作：婴儿骑红鱼，背生白翼，浮于深海；背景蓝绿渐变天空中飞满黄色小飞机。色彩浓烈对比，构图奇幻，融合童真与机械元素，营造梦幻而矛盾…
Illustration styles, analog media, colour
> **AI 视觉分析**: 画面为一位妆容夸张的女性肖像，风格暗黑哥特，色彩以灰白冷调为主，背景深邃。她红发盘起，眼妆浓重，皮肤苍白如尸，颈项修长，肩部露白纱。构图居中特写，突出其诡异而高贵的气质，充满戏剧张力。 Ac…

Try a question

Or pick one:

7-step pipeline

From an EPUB to a reader's tap, the full process is broken into 7 observable steps. Each owns its service, IO contract, and retry policy.

01
Parse EPUB
Uploaded to Supabase Storage; cheerio + jsdom extract chapter structure.
click-engine · book-ingest · cheerio + jsdom
02
Score sentences
Excerption Agent uses an LLM to find passages worth illustrating.
click-engine · excerpt · DeepSeek / Qwen
03
BookSignature
Signature Generator combines Qwen-VL with an LLM into an 11-dimension visual signature.
click-engine · signature · Qwen-VL + Qwen
04
Schedule generation
Image Generation Agent matches the signature to styles and dispatches tasks.
click-engine · visual-director · multi-tier
05
Bailian render
Calls Aliyun Bailian wan2.7 / qwen-image, references-driven.
aliyun bailian · wan2.7 / qwen-image
06
Anchor match
Hotspot Highlighter fuzzy-matches images back to chapter HTML positions.
click-engine · hotspot-render · fuzzy match
07
Render to reader
Interactive Reader renders the chapter with clickable hotspots.
interactive-reader · next.js 15

5-step agent decisions

The AI Studio conversational flow. Each step has an endpoint and an inspectable output.

01
Analyze
Understand source text and intent; locate subject and scene.
/api/agent/analyze
02
Corpus Search
Search the reference library and RAG entries for matches.
/api/agent/corpus-search
03
Select Style
Combine BookSignature with the matches to pick the run's aesthetic.
/api/agent/select-style
04
Enhance Prompt
Use RAG entries to fill in the prompt's visual language.
/api/enhance-prompt
05
Generate
Call Bailian; references drive consistency.
/api/agent/generate

One full agent decision log

Below: an internal record of one image task, showing each step's IO.

analyze
  input  : "她沿着河岸走，雪还在下"
  output : subject="a woman walking", scene="river bank, snow", mood="melancholy, quiet"

corpus_search
  hits   : "侧面光 / 逆光剪影", "冷调单色", "中长焦压缩"
  refs   : 3 张匹配参考图（来自 art_references 表）

select_style
  signature : era=1900s, palette=cold, abstraction=realistic, ...
  matched   : style_id=42 ("冬日故事")

enhance_prompt
  prompt  : "...low color temperature, side-light silhouette, mid-focal length..."

generate
  model  : qwen-image-edit-plus
  refs   : 2 张参考图
  output : 1 张 1024×1280 图

Taste as Method

Three reusable assets behind every image

Art reference library

Photography / painting RAG

Workflow + agent design

Art reference library

Photography / painting RAG

Light

Composition

Lens

Color

7-step pipeline

Parse EPUB

Score sentences

BookSignature

Schedule generation

Bailian render

Anchor match

Render to reader

5-step agent decisions

Analyze

Corpus Search

Select Style

Enhance Prompt

Generate

One full agent decision log

Now · pick a door

See

Make