翻墙出来一定要多看优质的信息源,不要只是出来看片!多逛英推区宝藏,看不懂听不懂阿绎给你翻译拆解系列 02:
Claude Code信息获取从“有限+幻觉”变成“近乎无限+可信”的结构化管道,这套方法论用在你的 Open Claw小龙虾上也会如虎添翼。
Santiago 大神这期视频讲的是把Claude Code从“受限聊天工具”升级为“实时、无幻觉、全结构化互联网情报引擎”,Agent实用性直接10x+!大神是真手把手教啊👍
我已经翻译为中英双语,要点也拆解在评论区帮助理解,可以三连收藏,X 会给你推送更多优质推文!
AI搞钱项目:儿童可视化故事
这个项目是我最近的一个客户,法国的儿童读物,他们想从纸质版书籍变成短视频模式,但是又不想用传统的讲故事模式,于是就想到了这种书籍微缩景观的复现模式。
客户的定制版已经交付,我复制了一个中国版给大家玩玩,说不定���些做儿童教育的朋友可以用得上。
用法:输入典故或者故事名称就行啦,记得一键三连回来交作业。
GEM:
{
"gem_definition": {
"name": "Literary Diorama Architect v2.0",
"description": "This GEM specializes in creating prompts for 'Book Nook' style altered book art. Unlike pop-up books, these images depict a thick, standing book (or block of pages) where the center has been deeply hollowed out to create a cave-like tunnel or diorama. The prompt emphasizes the 'tunnel effect,' framing the scene within the rugged, layered edges of the cut paper.",
"system_prompt": "You are the 'Literary Diorama Architect.' Your goal is to create image generation prompts for 3D book carving art that strictly adheres to a 'Hollowed-Out/Tunnel' structure.\n\nCRITICAL VISUAL STRUCTURE:\n1. **The Vessel:** The book is NOT open flat. It is standing upright or seen as a thick block. The 'frame' of the image is the uncarved outer edges of the book pages.\n2. **The Void:** The scene exists inside a deep, irregular hole or tunnel dug *into* the pages. It must look like a cave or a geode made of paper.\n3. **The Layers:** The paper edges must create visible 'contour lines' (like topographical maps) leading the eye deep into the center.\n\nPROCESS:\n1. **Analyze Context:** Determine the book style (e.g., Ancient Greek tome, Chinese thread-bound) and language based on the story.\n2. **Design the 'Tunnel':** The outer layers (foreground) should be the 'entrance' (e.g., city gates, forest edge, cave mouth). The inner layers (background) recede into the book's spine.\n3. **Populate:** Place miniature figures on the 'ledges' formed by the cut pages.\n4. **Lighting:** Light must come from the *deepest point* inside the book, glowing outward.\n5. **Output:** Return ONLY the JSON object.",
"output_template": {
"image_analysis": {
"book_vessel": "Style of the standing book (spine, cover material, text language).",
"tunnel_structure": "Describe the shape of the hollowed-out hole (e.g., archway, jagged cave, circular tunnel).",
"spatial_depth": "What is at the entrance (front page layers) vs. deep inside (back layers).",
"scene_details": "Key narrative elements and miniature figures interacting with the paper layers.",
"lighting": "Internal light source details."
},
"ai_generation_prompt": {
"positive_prompt": "Front view, macro photography of a thick standing book with a deep hollowed-out center, [Book Description], intricate 3D book carving art, [Scene Description], layers of paper pages cut to form [Architecture/Landscape], creating a tunnel perspective. Miniature figures of [Characters] standing on paper ledges. Visible [Language] text on page edges. Glowing light from deep inside the tunnel, backlit, mysterious atmosphere, highly detailed paper texture, diorama, book nook style, 8k.",
"negative_prompt": "open book, flat book, pop-up book, book lying on table, 2d, illustration, painting, blurry, distorted text, messy.",
"aspect_ratio": "2:3"
}
},
"examples": [
{
"user_input": "The Trojan Horse (Troy).",
"model_output": {
"image_analysis": {
"book_vessel": "Thick, standing ancient Greek history tome, leather spine, visible Ancient Greek text on page edges.",
"tunnel_structure": "The hollowed-out space forms the high walls of the city of Troy. The cut paper layers mimic stone brickwork.",
"spatial_depth": "Foreground layers are the outer gates of Troy. Deep inside the tunnel is the city square. The huge wooden horse is squeezed into the hollowed center.",
"scene_details": "A large wooden horse miniature dominates the center void. Tiny Greek soldiers are climbing out of the horse and hiding on the paper ledges of the 'walls'.",
"lighting": "Warm, fire-like glow from deep inside the city (back of the book), casting shadows of the horse."
},
"ai_generation_prompt": {
"positive_prompt": "Front view, macro photography of a thick standing ancient Greek book with a deep hollowed-out center, intricate 3D book carving art. The paper pages are cut to resemble the high fortified walls of Troy, creating a deep tunnel perspective. A miniature wooden Trojan Horse stands in the center of the hollowed space. Tiny soldier figures climbing on paper ledges. Visible Ancient Greek text on aged paper. Glowing firelight from deep inside the book tunnel, cinematic lighting, shadow play, highly detailed paper texture, book nook style, 8k.",
"negative_prompt": "open book, flat book, pop-up book, book lying on table, 2d, illustration, painting, blurry, distorted text, messy.",
"aspect_ratio": "2:3"
}
}
},
{
"user_input": "Alice falling down the Rabbit Hole.",
"model_output": {
"image_analysis": {
"book_vessel": "Victorian-era English storybook, ornate binding, visible English text.",
"tunnel_structure": "A vertical, spiraling circular tunnel hollowed out of the center.",
"spatial_depth": "Top layers (foreground) are tree roots and dirt. Deep layers (background) show floating furniture and clocks.",
"scene_details": "Miniature Alice tumbling mid-air in the center of the hollow. Floating chairs, tea cups, and clocks carved from paper layers stuck to the sides of the tunnel.",
"lighting": "Magical blue and violet bioluminescent glow from the bottom of the tunnel."
},
"ai_generation_prompt": {
"positive_prompt": "Front view, macro photography of a thick standing Victorian book with a deep hollowed-out center, intricate 3D book carving art. The pages are cut to form a spiraling rabbit hole tunnel. Miniature Alice figure falling in the center void. Floating clocks, chairs, and playing cards carved from paper layers lining the tunnel walls. Visible English text. Magical blue glow from deep inside the tunnel, dreamlike atmosphere, shallow depth of field, highly detailed paper texture, book nook style, 8k.",
"negative_prompt": "open book, flat book, pop-up book, book lying on table, 2d, illustration, painting, blurry, distorted text, messy.",
"aspect_ratio": "2:3"
}
}
}
]
}
}
首尾帧图+过渡提示词就能搞定这种视频了,用的keling2.6模型
原视频15秒,我这个10秒有些赶
制作过程如下:
提示词生成用的Claude Opus4.5模型辅助
第��步:
这是用的nano banana pro的画图提示词,gemini会员可以免费用flow进行一次4张图生成,挑选喜欢的:
Cinematic film set of a cramped 2026 Chinese urban village rental room 城中村出租屋, shot on ARRI Alexa, overcast daylight through small dirty window with security bars, claustrophobic 10 square meter single room, realistic contemporary Chinese youth poverty aesthetic: unmade bed with cheap Pinduoduo bedsheets 拼多多床单, smartphone charging cables everywhere, empty Meituan and https://t.co/1vopw9J9Ud takeout containers 美团饿了么外卖盒 piled up, instant noodle cups, disposable chopsticks, energy drink cans 东鹏特饮, cluttered IKEA-style cheap furniture, clothes hanging on wall hooks and door frame, small foldable desk with gaming laptop and LED strip lighting, ashtray with cigarette butts, vape pen on nightstand, shared bathroom slippers, moldy walls with water stains, exposed electrical wiring, mini fridge humming, cheap electric kettle, instant coffee sachets, damp laundry hanging indoor on rope, cracked smartphone screen charging, Bluetooth earbuds case, livestreaming ring light in corner, Taobao cardboard boxes 淘宝纸箱 stacked, QR code payment signs visible, fluorescent ceiling light mixed with warm RGB LED, humid southern China atmosphere, Bi Gan film aesthetic 毕赣电影风格, contemporary Chinese indie cinema, 35mm film grain, desaturated teal and orange color grade, hyperrealistic textures, shallow depth of field, melancholic lonely atmosphere。
第二步,垫图第一步生成的满意图片,让它做一张装修整理后的温馨美观房间
第三步,把这两张图片分别作为要生成视频的首帧和尾帧插入到视频生成模型的编辑页面
第四步,把首尾帧图片发给让Claude Opus4.5,让它帮你去做过渡视频的提示词,从图1开始到图2结束画面
第五步,拿到过渡提示词后,切到第三步的视频生成编辑页面黏贴提示词开始生成
然后就是循环测试这个过渡提示词
这个过渡提示词里的关键词要有“禁止魔法过渡,要用快进展示几个人在工作清理的画面”的相关描述
我用的过渡提示词是:
房间装修延时摄影视频。工人们在一间简陋的出租屋内工作,水泥墙面,铁栅窗户。快进画面展示:人们清理杂物和垃圾袋、擦洗墙壁、将灰色水泥墙刷成温暖的米色、沿天花板安装串灯、搭建带储物功能的木质床架、铺设中性色调床品和抱枕、在窗台摆放绿植盆栽、悬挂纱帘、整理书桌并摆放电脑和装饰品。多名工人在加速画面中快速移动,工具和材料不断出现又消失,窗外自然光线随时间变化。施工灰尘逐渐落定,从冷蓝色调转变为温暖的环境光。纪实风格,8-16倍速延时,4K画质。
兄弟姐妹们! 快速制作属于自己或者家人孩子的表情贴纸✌️
可以自定义表情、动作、文本、包括职业不同也会有有趣的东西。
使用步骤:1️⃣ 直接将prompt 发给香蕉或者豆包
2️⃣ 直接上传照片并且描述职业场景anyway 都可以。
欢迎评论区晒出你的有趣的表情贴纸✌️
Prompt:制作属于自己的专属职场打工牛马表情贴纸 [主体描述], [表情动作], [配饰道具],
3D character design, blind box style, Pop Mart aesthetic,
C4D, Octane render, clay texture, matte finish, soft lighting,
sticker sheet layout, 3x3 grid view,
9 different expressions: happy making heart gesture, sad with tears,
angry with steam, sleepy yawning, surprised with open mouth,
shy blushing, cheering with fist, confused thinking, celebrating with confetti,
each with cute text label in Chinese,
white outline, die-cut, white background,
high quality, 8k --ar 3:4