什麼?這動物圖片可以上國家地理?

flydean發表於2024-07-09

stable difussion中大部分的模型都是關於人的,今天交給大家一些不一樣的:如何生成動物圖片。在這篇文章中我們將會學到如何生成逼真的動物,可愛的動物,還有幻想中的動物。

準備工作

當然前提是你需要一個SD的軟體,你可以用本地的SD webUI或者雲端的SD環境,全都是可以的。

真實的動物

由於目標是生成逼真的攝影影像,因此你需要包含關鍵字“photo”。

提示應該像這樣開始

photo of

主題

首先,你需要選擇你的主題。例如:

Dog 狗,Cat 貓,Pig 豬,Lion 獅子,Deer 鹿,Monkey 猴子,Elephant 大象,Horse 馬,Bear 熊

場景

場景控制背景和周圍環境。由於關聯效應,如果不新增場景關鍵字,通常會得到野生動物的自然棲息地。

森林, Forest;海灘, Beach;山脈, Mountain Range;河流, River;沙漠, Desert;公園, Park;田野, Field;城堡, Castle;宮殿, Palace;冰川, Glacier;瀑布, Falls;湖泊, Lake;花園, Garden;洞穴, Cave;雪地, Snowfield;火山, Volcano;熱帶雨林, Tropical Rainforest;草原, Prairie;市區, Urban Area;鄉村, Countryside

燈光

燈光對影像的外觀有很大影響。良好的燈光使影像變得有趣。

氛圍燈, Mood Lighting;情緒燈, Moody Lighting;工作室燈光, Studio Lighting;壁角燈, Cove Lighting;柔和照明, Soft Lighting;硬朗照明, Hard Lighting;倫勃朗燈光, Rembrandt Lighting;體積光, Volumetric Lighting;生物光, Bioluminescence;電影燈光, Cinematic Lighting;雙色燈光, Bisexual Lighting;逆光, Back Lighting

其他

使用類似於生成真實人物的關鍵字。例如:

dslr,ultra quality,film grain,8K UHD

當然,提示詞並不是越多越好,使用過多可能會導致解剖結構不佳。

下面的一些關鍵詞可以增加野生動物的美感:

  • National Geographic Wildlife photo of the year
  • The American Landscape Contest
  • Wildlife photography contest

模型

我們應該選擇一些真實的模型,比如:

  • Realistic Vision
  • Dreamlike Photoreal

真實動物影像示例

以下是一些用於生成逼真影像的示例提示。隨意使用或重新混合。

模型: Realistic Vision v60B1

提示:

National Geographic Wildlife photo of the year, photograph, Dutch angle shot of a Giant (tiger:1.1) , complex Low Contrast background, Fall, Very wide view, Contemporary, Hopeless, masterpiece, Superflat, Side lighting, still, Kodak Tri-X 400, F/14, Folded Wires, ethereal magical atmosphere

否定提示:

spring, painting, abstract, camera, anime, cartoon, sketch, 3d, render, illustration, simplistic, minimalism, plain, simple, modern, ordinary, mundane, cropped, worst quality, low quality, poorly drawn, low resolution,

image-20240423184010614

模型: Realistic Vision v60B1

提示:

National Geographic Wildlife photo of the year, photograph, extreme wide shot of a (panda of Past:1.1) with Crow skin, Deeply Thoughtful, Raining, Bokeh, Manic, dtx, rim light, Orton effect, Phase One XF IQ4 150MP, F/5, BW, fairy tale, rainbow swirl, beautiful

否定提示:

modern, ordinary, mundane, painting, abstract, camera, anime, cartoon, sketch, 3d, render, illustration, sharp, focus, ugly, gritty,

image-20240423185019232

提示:

National Geographic Wildlife photo of the year, photograph, Fluffy elephant, Sharp and in focus, Tranquil, Breakcore, Sunlight, High Shutter Speed, dslr, Selective focus, Technicolor, HDR, RAW photo, realism

反向提示:

Negative prompt: painting, (abstract:1.3), camera, anime, cartoon, sketch, 3d, render, illustration, bokeh, blurry, blur, simple, plain, unrealistic, impressionistic, low resolution, surrealism,

image-20240423185708615

可愛的動物

簡單描述

如果你只是想生成一些可愛的動物圖片,那麼包含“cute”一詞的非常簡單的提示就可以了。選擇一個模型來實現某種風格。

型號: DreamShaper

提示:

a cute cat

image-20240423200011196

更進一步

你還可以新增關鍵字以使用相同的模型進一步修改樣式。

提示:

photograph, crowded blossoms, intricate details, side view shot of a (a cute cat:1.3) - British Shorthair Cat hybrid, simple background, Sharp and in focus, Cel shaded, Fantasy, exquisite, moody lighting, Cinematic, Ilford HP5, F/5, Calotype, (electric indigo and dark yellow:0.7) , Pixabay

否定提示:

simplistic, minimalism, plain, cropped, worst quality, low quality, poorly drawn, low resolution, painting, abstract, camera, anime, cartoon, sketch, 3d, render, illustration, gothic, monochromatic, ugly, gritty, bright, sunny, light, vibrant, colorful, industrial, mechanical, geometric patterns, bokeh, blurry, blur, surrealism, ornate, complicated, highly detailed, cluttered, disordered, messy, noisy, maximalism, realism, photorealism, modern, ordinary, mundane,

image-20240423195855028

向量風格

如果想要繪製向量風格的圖片,那麼可以加上關鍵字vector

Prompt:

a Intellectual wolf, whole body, Anthropomorphic, portrait, highly detailed, colorful, illustration, smooth and clean vector curves, no jagged lines, vector art, smooth

反向提示詞:

cluttered, disordered, messy, noisy, full body shot, photograph, photorealism, simplistic, minimalism, plain, simple, dark, moody, night, monochromatic, washed out, realism, frame, watermark,

image-20240423210609479

更特別的圖片

除了使用基本的描述詞之外,你還可以使用一些lora來創造特別樣子的動物圖片。

比如這種可愛的擬人風格。

lora:Round world chubby little animals doll toys

提示詞:

Fat cute wolf,with a belly,in the forest,snow,<lora:20240130-1706578549277:1>,

image-20240423203124063

還有做成麵包形狀的動物:

Lora:CSTB v1

提示詞:

<lora:CSTB_v1:0.6>,lying,food,no humans,animal,looking up,painting (medium),animal focus,bread,food focus,still life,wolf,

image-20240423204327431

機械風的動物:

Lora: Animal Mecha

提示詞:

best quality,masterpiece,highres,original,extremely detailed wallpaper,perfect lighting,extremely detailed CG,<lora:add_detail:0.3>,perfect_photography.txt,<lora:Animal_mecha:0.8>,animal mecha,cyborg,wolf face,

image-20240423205132808

人形風格的:

Lora:animal modal

提示詞:

((low contrast)),animalrizz ,<lora:animal_model:0.5>, ((((wolf)))) (Captain America:1.3),Blonde hair,solo focus,muscular,(((Blue suit))),Helmet,Blue eyes,Avengers logo,Cityscape,Super soldier,Patriotism,{{{Master piece:1.3}}},{{{realistic}}},{{cinematic lighting}},{{{perfect anatomy:1.3}}, handsome man, {{extremely detailed}}},long eyelashes,{{perfect face}},{{perfect eyes}} refined,{soft lighting},{{cool}},(detailed body),(Masterpiece)),(((best quality))),( full body:1.4),good anatomy,short hair,heroic features,tall,fit,large chest,((highly detailed features)),(((very detailed face))),noble features,intricate,(depth of field),dynamic lighting,highly detailed body,detailed features,4k,textured,1 man,looking at viewer,Captain America logo,Star on his chest,Red and white stripes on his suit,Dark night sky,Moonlight,New York City,Heroic,Brave,

image-20240423211816225

寫在最後

現在你已經熟悉了一些生成動物圖片的技術,你可以使用這些技術在Stable Diffusion中生成你想要的動物影像.

但是任何技術生成的影像都可能是不完美的,所以你需要做的就是使用img2img來進行進一步的修改。

點我檢視更多精彩內容:www.flydean.com

相關文章