Googleが発表したWhiskが生成AIのことがよくわからない人も、プロンプトが書けちゃう人でも楽しいことに気づいて、一瞬でハマってしまいました。画像生成速度も速くて、サクサク作れちゃうので、ついついあれこれ作っちゃってます。

Google Whiskとは
Geminiに聞いてみました。
Q. Google Whiskとはなんですか?
300文字以内で中学生でもわかるように教えて下さい。
Google Whiskは、Googleが開発した画像生成AIツールです。
普通、AIで絵を作る時は言葉で指示を出すけど、Whiskは複数の写真を選んで「混ぜて」と指示を出すだけで、AIが新しい絵を作ってくれます。
例えば、猫の写真と夕焼けの写真を選んで混ぜると、夕焼けを背景にした猫の絵ができます。まるで泡立て器(Whisk)で材料を混ぜるみたいだから、この名前がついたんだって。
Whiskは無料で使えるから、みんなも試してみてね!


詳しい情報は以下のサイトあたりを見てもらうと良さそうです。



やってみたこと
お昼休みにスマホでポチポチ。
テンプレートがあってモデルの画像をスマホから選ぶだけでできちゃいます。

雪山の写真はプロンプトで作りました。


そしてすごいのはプロンプトでの再現性があること。
例えば生成したこの画像。

なんと、下のプロンプトで再現できちゃいます。
A grainy, 1960s film aesthetic image depicting a person with light skin tone, wearing a black hoodie and blue jeans, sitting inside a kotatsu.
The person is in the process of putting on a gold and red mask.
The kotatsu is partially visible, suggesting a warm, cozy interior contrasting with the cold exterior.
The background is a snow-covered landscape with rolling hills densely covered in coniferous and deciduous trees, rendered in muted whites, grays, and browns.
The sky is a uniformly light gray.
The image features high contrast, deep blacks, and bright whites, with a noticeable film grain and a warm, yellowish tint.
A slight vignette darkens the edges, and the focus is sharper on the person in the kotatsu, softening towards the periphery.
The lighting mimics a single, powerful flash, creating dramatic shadows.
The overall impression is a nostalgic, slightly imperfect, and cool aesthetic, capturing a candid moment in a cold, quiet winter scene.
The scene is asymmetrical, emphasizing a sense of captured spontaneity.




プロンプトをちょっといじってあげるとこの通り。2020年代風で金色のマスクにしてみました。

いろいろ作ってみた
そんなわけで、いろいろ作ってみました。
自分が持っている写真をもとにいろいろ作れちゃいます。楽しい!
もちろん失敗も多数あります
特に2頭身キャラを車に乗せるのはまだまだ難しいようです。
プロンプトでえっちらおっちらやってました。

その他生成した画像ギャラリー
細かく見ると完璧じゃないけど、画像をかけあわせて生成してくれるのはとても良いです。







GoogleのWhiskは、これまでのGoogleの画像解析技術と固有名詞の紐づけが高レベルで活用されてると感じました。
興味がある方、お試しあれ!



コメント