你出題，我叫AI畫畫

Melody on a lovely grass plain with long horizon

香港打中國

Fight club

Whisky, Brandy and Rum

世界末日 回到過去

鯨魚 河馬 毛蟲 女機械人

Cloud, six, love

Chinese invade Taiwan results in epic failure

凌晨時份，John Depp 一個人喺大排檔食豆腐火腩飯

第二張個feel勁Hilary Hahn

Thalassophobia

https://github.com/lucidrains/imagen-pytorch
已經有人replicate咗個architecture
而家只係等緊第一個pretrained model

Are there any pretrained models yet?
Not at the moment but one will likely be trained and open sourced within the year, if not sooner. If you would like to participate, you can join the community of artificial neural network trainers at Laion (discord link is in the Readme above) and start collaborating.

上帝的打賭

自己train 要幾耐？
租多啲電腦train 會快啲？

認唔認到德文
Hier mag ich, hier bleib ich.

三隻貓

肥婆 美極 自助餐

tank driving on tiananmen square

Pistol, Scissors, Blue moon, Ponytail

伏尼契手稿d friend

唔知 simon stalenhag 會點諗

樓豬今晚嘅心情 : In this turbulent occasion, a man continue to work with his computer silently, sci-fi painting by Simon Stalenhag



Wallpaper嗰度今日整咗俾你 :


其實用Dall E 可以整到，只要將其中一部份遮住咗，然後再重新俾個AI畫畫出嚟, 就可以產生一個新嘅部份然後再駁埋一齊，如此類推 :





另外其實根據 Dall E Prompt Book, 生產到嘅圖片可以分為：
1. 插畫 (Illustration)
2. 相片 (Photography)

今日試咗少少關於相片嘅操作，可以同大家分享吓。 細佬仲特登請咗兩位由日本同韓國嚟嘅姐姐做Model :


第一位姐姐係 : Portrait of beautiful young Korean female idol wearing camisole, in 2020 fashion pink style, golden hour in manhattan, Zeiss 150mm f2.8 Hasselblad, award-winning photo

第二位姐姐係 : Portrait of beautiful young Japanese female idol wearing camisole, in 2020 fashion pink style, golden hour in manhattan, Zeiss 150mm f2.8 Hasselblad, award-winning photo

我我哋可以將題目拆解：
a. Portrait = 肖像圖
b. Beautiful young Japanese female idol wearing camisole = 主要描述/場景/ 物件年份
c. in 2020 = 年份
d. fashion pink style = 色調
e. golden hour in manhattan = 燈光場景
f. Zeiss 150mm f2.8 Hasselblad = 拍攝鏡頭/ 光圈
g. award-winning photo = 整體風格/ 感覺

所謂只要有心人人都可以做食神; 同樣地，只要有心人人都變成大攝影師大畫家

啱啱發佈成果
未有得俾你玩住
Google Parti 啲圖冇咁 P 圖 feel