生成AI

０５：生成AI

※背景の画像はDALL-Eで生成しました

2024年7月1日

ChatGPT4o(DALL-E)で生成

「プロンプト」の列の右端にある「 V 」を押すと追加の情報を見ることができます。

プロンプト

出力結果

私は大学生で、情報工学(特にプログラミングとゲーム開発)を趣味としていて、プログラミングによる自動生成が好きです。16bit pixcel artで私のポートレートを生成してください。ただし、人物像を描いてはいけません。

A 16-bit pixel art style portrait representing a university student interested in programming and game development, but without depicting a person. The scene includes a computer with code on the screen, game development tools, and elements of procedural generation like random terrain or character sprites. The background is filled with books, posters of classic games, and other items related to information engineering and game development.

情報量が多すぎるので、もっとシンプルに

A 16-bit pixel art style representation of a university student interested in programming and game development. The scene includes a computer with code on the screen and a simple background with elements like books and game development tools. Keep the design minimalistic.

文字はいりません。また、人物をメインにしてください。

A 16-bit pixel art style representation of a university student interested in programming and game development. The scene should feature a student sitting at a desk with a computer, surrounded by simple elements like books and game development tools. The design should be minimalistic and focused on the character without any text.

ぽーとれーとにしてください

A 16-bit pixel art style portrait of a university student who is interested in programming and game development. The portrait should focus on the upper body and face of the student, depicting them with a calm and focused expression. The background can include simple elements like a computer or books to indicate their interests but should remain minimalistic.

バグって色んな顔とか人が混ざった感じにして

A 16-bit pixel art style portrait of a university student interested in programming and game development, but with a glitched appearance. The face should have various features from different faces mixed together, creating a surreal and fragmented look. The background remains minimalistic with elements like a computer or books, indicating their interests.

背景を有機的にして(オメガフラウィみたいに)

A 16-bit pixel art style portrait of a university student interested in programming and game development, but with a glitched appearance. The face should have various features from different faces mixed together, creating a surreal and fragmented look. The background should be organic and surreal, similar to Omega Flowey, with twisted and bizarre elements that contrast with the character.

▼ オメガフラウィ(PhotoshopFlowey)

https://undertale.fandom.com/wiki/Photoshop_Flowey/In_Battle より

UNDERTALEに出てくるボス。フォトコラージュで作られている。

有機的で、生物や植物などの写真のフォトコラージュ風に

A 16-bit pixel art style portrait of a university student interested in programming and game development, but with a glitched appearance. The face should have various features from different faces mixed together, creating a surreal and fragmented look. The background should be a photo-collage style, incorporating organic elements like plants and creatures, blending pixel art with realistic images to create a chaotic and surreal atmosphere.

輪郭線をドット絵と写真の混ざりあった感じに

A 16-bit pixel art style portrait of a university student interested in programming and game development, but with a glitched appearance. The face should have various features from different faces mixed together, creating a surreal and fragmented look. The background should be a photo-collage style, incorporating organic elements like plants and creatures, blending pixel art with realistic images to create a chaotic and surreal atmosphere. The outlines should be a mix of pixel art and realistic photo elements, creating a blended, hybrid look.

プロンプトを全て無視して、様々な写真を乱雑に配置してカオス化

A chaotic photo-collage style image with various photographs placed randomly and in a cluttered manner. The image should include a mix of plants, flowers, hands, fingers, eyes, and other random elements, creating a sense of chaos and disorder. The collage should appear as a haphazard assembly of different photos, with no regard for order or context, resulting in a visually overwhelming and fragmented artwork.

生成AIによる画像生成は、思い通りの絵を描けない人やコストを削減したい人にとっては素晴らしいツールです。（私もキャラデザをAIにある程度任せたりしてます）ただ一方で、AIに作ってもらうだけでは「作品を作る実感」をあまり感じられないのもすごく分かります。

これは、AIを使っているからというより、AIの出力に受動的すぎるのが原因かもしれないと個人的に考えています。出力をそのまま受け入れるだけではなく、AIに描かせた画像の一部を更にAIで上描きしたり、プロンプトを変えてみたり、自分の手で編集してみたり、自分の作品をAIに編集してもらったり、FineTuningしてみたり...。

そういったことをすると、作品に対して思い入れが出てきて、作品を作っているという実感が湧いてくるのかなと考えたりします。