見出し画像

Midjourneyでリアル系SF画像を作成してみる プロンプトサンプル20個位あります

本記事はMidjourneyを使い始めた人が対象となっています。とりあえずdiscordに加入したが, プロンプトが良くわからない, でもSF映画で出てくるような綺麗なワンシーンを簡単に作ってみたいという人におすすめです。

機械仕掛けの鳩 Midjouneryも以前描いたと言われる
Create a photorealistic image of an original logo featuring a mechanical pigeon with elements of pink neon lighting. The design should convey innovation and creativity, suitable for a technology or design company. The pigeon should appear as if it could exist in the real world, with detailed gears and metallic parts, illuminated by vibrant pink neon accents in a realistic setting. The background should be minimalistic to ensure the logo stands out clearly.

自分なりにうまくいったプロンプト紹介していきます。例にもれず, 前置き, 基本的な使い方, 蛇足, 独り言などが長いので目次から本編 (6章 制作例 プロンプト例) へリンクすることをお勧めします。


0. はじめに

2023年は生成AIの年としばしば言われています。DALLE3が登場してから面白さにとりつかれ画像生成AIで連日遊んでいます。画像生成にとどまらず, 短編動画にも手をだしたりもしました。AIのおかげで今まで素人には全く手がでなかった動画作成が簡単なものであれば, テキストプロンプトのみで作成可能となっています。

まさに底なし沼状態です。

DALL-E3 AIの沼にはまる

前回, Runway短編動画を作った際に感じたのが, やはり画像クオリティの限界でした。DALLE3やMidjouneryと比べ, 表現力が劣る印象が否めなかったです。Midjourney, Stable Diffusionは玄人感があり, 使いこなすのが難しそうなイメージがあり中々手がでませんでしたが, やはり綺麗な画像を作りたいと思い今回挑戦してみました。

Runwayのテキストプロンプトで出力された画像

画像生成AIの中でMidjouneryが選ばれている大きな理由として高画質表現力の高さということです。DALL-E3では特定の作品の使用キャラに似ている画像の作成には強い制限がかかります。

一方, Midjouneyでは『マ●オ』というワードを出さなくても, 連想するような単語で構成されたプロンプトで容易に生成されます。

Midjourney:マリ●酷似というかマ●オそのもの!?

知っていても敢えて作らないDALLEと作品名を指定しなくても勝手にというますか, 模倣作品がつくれてしまうMidjourneyは対照的です。

その他, DALLEとの大きな違いとして, Midjouneyの画像生成はDiscordchatルームで作成します。

他のuserも同じセッションで画像作成していますので眺めているだけで楽しいですし, 使い始めは特に他の人のプロンプトが大変参考になります。

1. 著作権について

画像生成AIを使うときの最も注意すべき点として他の作品, 著作物に対する著作権侵害となります。このあたりの法整備は現在進行形で結論がでていませんので, 画像生成AIを使った生成物で商用利用するなどは, 使用するuser側自己判断が重要となってきます。

最近の動きとして作家Open AIとの訴訟ではOpen AI側が優勢となっています。この流れは画像生成AI界隈に追い風になるかもしれません。実際使ってみてビックリしたのが初めてSFの宇宙人を作成したときです。

いわゆる『エイリアン』映画エイリアン (©20世紀フォックス)を避けるために『眼と角』を持つという表現をプロンプトに入れましたが, 出てきたのは以下です。

Midjouneryで『眼と角』をもつ異星人クリーチャー

後で検証しましたが, 『角』『目』というワードが入っていても『宇宙船』『黒』というワードがプロンプト内に入っているため"映画エイリアン"を想起する強力なワードのため, 事前学習しているいわゆるエイリアンを呼びよせてしまったらしいです。

以下は©20世紀フォックス有名映画『インデペンデンスデイ』の名シーンのリスペクト画像です。『インデペンデンスデイのホワイトハウス爆破シーン』とだけプロンプトしたものです。

実際の映画のシーンとは違うものが出力されました。次にシーンの詳細描写をChatGPTで英訳し, そのプロンプトを用いてMidjourneyで生成してもらいました。

こちらは『インデペンデンスデイ』入れなくても, 実際の映画に近いものが生成されました。生成AI特性上, もちろん全く同じシーンの作成は基本的には不可能でリスペクトシーンとなっています。

『Independence Day』をプロンプトに含めていない

一方『Independence Day』という単語を含めた場合, 多少原作近い?ような気がしないわけでもないですが大差ない結果となっています。

Independence day というワードを含めなくても, そのシーンの状況説明詳細プロンプトしたら結果的に『Independence Day』のシーンが再現されました。

Independence day という単語が無くても同じような出力

というわけで以上から何が言いたいかというと, user側意図していなくても, また, オリジナルと思って作ったキャラクターであっても, どこかの作品にすでにでていて, さらに酷似したものであると, それを利用した商品著作権侵害になってしまうことがあるということです。

GPTにこれまでのやりとりをまとめてもらいました

・映画作品名をプロンプトに入れることについて
映画〇〇のシーンとプロンプトする方が簡単に名シーンを作ることができます。以下のシーンはいろいろなSFシーンを映画のシーンで作ったものです。

火星に降り立つ宇宙服の人

Hyper-realistic photograph of a movie scene from ■he M■r■i■n, two astronauts in spacesuits up close on Mars surface --ar 16:9

ロボットの戦闘シーン

Hyper-realistic photograph of a movie scene from 'All ■ Need Is ■' featuring robots engaged in combat, showcasing detailed designs and dynamic action in a futuristic battlefield setting --ar 16:9 

こちらの記事参考にさせていただきました。

宇宙ステーションからの地球のシーン

Hyper-realistic movie scene of z■r■ g■a■ity, International Space Station with Earth in the background --ar 16:9 

・芸術家の名前をプロンプトに入れることについて

一般的な人物画の場合, 微妙な特徴を日本語で表現することは難しいので, 例えば以下の画像は芸術家, 画家の名前をプロンプトする以外には出力困難と思われます。 

girl in the robotic suit art by ad■■ vt d■■, in the style of shallow depth of field, elegant, emotive faces, hyper-realistic water, s■■h■■o m■■uo, light black and azure, mi■■■el kom■■■k, loose and fluid --ar 16:9 --v 6.0 --style raw --s 250

映画作品名をプロンプトに入れ, 出力された画像を使って商品をつくることは悪質と言えます。しかし, 状況設定が特殊なシーンや特徴的なアニメなどのキャラクターは作品名を入れなくても簡単に同じようなものがつくれてします。

画像生成AIで著作権侵害をしていまう

繰り返しなりますが, 悪意がなくてもたまたま酷似するものが作れてしまいます。というわけで本質的には作品名を入れないことは著作権侵害の回避には全くならないと思われます。

それでもuser側に出来る事と言えば

・敢えて作品名, 人物名を入れない
オリジナルシーン (と自分が考えている) を作成するように努める
・映画のワンシーン, 特に有名シーンと酷似するものが出来てしまったら
 盗作と判定されるリスクが高いので用いない。

くらいでしょうか。

2. 良く使うコマンド

良く使うコマンド, 便利なコマンドを挙げてみます。こちらの記事基本的な使い方を学びました。

コマンドは全部で20種類以上あります。

2.1 /Info

アカウント情報 (主に使用履歴) を見るコマンドです。

chat欄に『/info』と打ち込み『Enter』を押します。
① 現在加入している料金Plan (Subscription), ② Fast modeで高速に生成できる残り時間, ③ Lifetime Usage これまで作成した枚数が確認できます。

② 後, どれくらいFast Timeが使えるかを確認します。 14.65/15 hours (97.67%) であればあと14.65時間残っているという意味です。時間は画像生成につかう正味の時間のことで1枚生成すれば大体1分消費されます。

③ 作成枚数 これまで作った累計枚数が1000枚を超えると良い事があります。

Midjourney AlphaのWeb版が使えるとのことです。Discord上で作動するのではなく専用のWebアプリで使え, いろいろと便利機能, 使い勝手が向上しそうです。

2.2 /settings

設定 (Confing) を確認, 有効にするコマンドです。

『/settings』+『Enter』で下の画面が出てきます。現在有効になっている設定が表示されています。有効なボタンは緑色表示されており, 例えば現在, Model V6 (最新のモデル), Stylize med, Public modeなどが有効になっています。

以降で説明しますプロンプトでも同じような設定で作成することができます。例) RAW mode→プロンプト末尾に 『--style raw 』を加える。

こちらの記事にsettingsの詳細が分かりやすく説明されています。

2.3 /describe

画像を見せてプロンプトを作ってもらうコマンドです。GPT4V (ision) に近い機能かもしれません。書きたい手本の画像があり, プロンプトを知りたい場合, 似た画像を描きたいときに使います。

Midjourneyプロンプト作成能力を調べるために, DALLE3に『AIの画像解析能力を測るため』の表現が複雑な絵をまずかいてもらいました。

Create a highly detailed, realistic painting that challenges the AI's ability to interpret and illustrate complex prompts. The image should depict a scene that is intricate and nuanced, with multiple layers of meaning and interpretation. It could involve abstract concepts, surreal elements, or a combination of both, set in a realistic environment. The artwork should push the boundaries of AI's understanding, requiring a deep level of analysis to grasp the full extent of the scene. The composition should be rich in detail, with a focus on realism, to truly test the AI's capability in creating complex and thought-provoking imagery.

chat欄に『/describe』と打ちます。

上に 『image』『link』を選ぶウインドウが出てきますので『image』を選びます。

画像のurlを持っている場合は『link』からもいけそうです。
先ほどの抽象画の画像ファイルを出てきたウインドウにドラッグアンドドロップします。

『enter』を押ししばらく待ちます。プロンプトの候補が4つ出てきました。

『1番』を選択してみます。※Midjouney情報が送信されますので個人情報を含まない画像にしてください。

なかなかの再現度だと思います。

架空SFシーンをChatGPTに描いてもらいました。

A highly imaginative and complex illustration of a man with his head set on a wheel, wearing headphones, in a style that combines elements of a tesseract, high-speed film, circuitry, medicalcore, cobra, and mirrored imagery, with a strong emphasis on science-based aesthetics. The composition should merge these diverse influences seamlessly, creating a surreal and captivating visual experience. The man's pose and expression should convey a sense of motion and innovation, as if he's navigating through a futuristic landscape filled with advanced technology and abstract concepts. The use of mirrored elements should enhance the complexity and depth of the image, making it a rich tapestry of modern and speculative ideas.

素材のイメージを残しつつ芸術が加えられた印象です。

実写系も渡してみます。Runwayで適当につくった架空のSFシーンです。

Runway :SFシーン

全然違うのも含まれています。

似たような画像を作成します。ChatGPTのGPT4V画像を渡し『再現プロンプト』を考えてもらいました。

A female figure stands on an elevated platform overlooking a sprawling futuristic city at night, illuminated by thousands of lights. In the foreground, a massive, detailed starship dominates the view, its design both intricate and industrial. The city's architecture melds modern with futuristic, featuring high-rise buildings with glowing windows and streets bustling with activity. The mood is atmospheric and tense, as if the woman is witnessing the quiet before a storm in this advanced civilization. --ar 16:9 --v 6.0 --style raw

色使い, スタイル (画風 artistic style) はオリジナルと異なりますが, キャラ, オブジェクトの配置, 構図は再現性が高いです。次にMidjouneyのdescribeコマンドです。

スタイル (流派, 画家) について具体的な提案がありました。1番のサンフランシスコ=ルネサンスを選んでみます。※現役イラストレーター漫画家などがいれば取り扱い注意です。本記事ではあくまで再現実験として行っています。

ChatGPTよりはスタイルの再現性が高いですが, 位置が微妙です。というわけで再度, MidjouneyにもらったプロンプトをChatGPTに微調整してもらいました。

ビルから見下ろすように変更

元の画像とスタイル, オブジェクトの相対位置などが再現されたものができました。

A futuristic female atop a high-rise building gazes down at a colossal starship in a city resembling a dark gray and dark bronze future San Francisco. The scene is a still from a movie, capturing the essence of a renaissance period with rough clusters of urban structures and meticulous military details. Translucent planes glide through the backdrop of intense action scenes. The overall style is raw, highlighting the unrefined textures and the stark contrast between the city's architecture and the advanced technology of the starship. --ar 16:9

まとめとしてMidjouneyの『/describe』コマンドでプロンプトをもらい, ChatGPTに細かなニュアンスを微調整してもらいます。

全く同じものが作りたい場合は 4章イメージプロンプトとを参考にしてください。

3. パラメーター設定:末尾の追加プロンプト

 Midjourneyはプロンプト末尾をハイフンハイフンで区切り『-- ■』■に追加の設定を指定し出力画像をコントロールする仕様です。以下が公式のパラメーターリストです。

先ほどの宇宙船のシーンのプロンプトの末尾に種々の条件を追加しどのような変化があるか調べてみます。基本画像は以下の画像です。アスペクト比16:9, modelは最新のMidjourney V6.0です。

『--ar 16:9』などのパラメーターを末尾に連続して入れると反映されないことがありますので『半角スペース』を入れるようにしています。

Hyper-realistic photograph of a movie scene, looking up at a colossal spaceship from a low position over a city --ar 16:9 --v 6.0 

3.1 --ar

アスペクト比 サイズ比 (縦と横の長さの比率) のことです。プロンプトの「最後に『 --ar 4:3』を追加すれば縦3対横4の比率に変更されます。映画と相性の良い比率は16:9と思いますので以降は16:9で固定します。

3.2 --stylize

Stylizeの変更

プロンプト末尾に『 --s (stylizeの値 0-1000)』を付け加えるか, settingsのボタンでStylizeを変更することで設定できます。stylizeは創造性, 芸術性の高さを意味します。

高い芸術的, プロンプトで書かれていないことも表現されます。逆に低い芸術性低く, プロンプトに忠実になります。初期設定はStylize mid, 100に設定されています。

ボタン                                     stylizeの値
Stylize low                                50
Stylize mid (デフォルト)          100
Stylize high                               250
Stylize very high                       750

aitechworld.info
s:0

確かにプロンプト忠実 (宇宙船とビルの写真としか言っていないので..) で写実的です。

low: s 50
middle (default): 100
high: s 250
very high: s 750
s 1000

sの値が増えるにしたがって創造性, 芸術性が高くなっているのが分かります。

3.3 --style raw

Rawmode のonとoff

直訳するとraw なま 『生』です。style rawオンにする (Raw modeを有効) とMidjourneyによるデフォルト加工がなく, 素材 (=プロンプト) に忠実, 写実的な絵が生まれます。

RAW Modeが有効になっている

/settingsコマンドでRAW Modeのボタンを押すか。プロンプトの末尾に『 --style raw』を付け加えると有効になります。

リアルなSF画像ではリアリズムとディテールを強化するため原則, style raw onが相性良いとのことです。style rawをonにしつつ微調整をstylize値で行うことが推奨されています (ChatGPT的に)。理由を考察してみます。

style raw  onでは自然, 写実的, 一方, offではMidjouneyによって芸術的加工が加えられ, 抽象的になります。基本画像でrawon/offの比較を検証します。元となる画像offです。

middle (default): 100
raw off
Hyper-realistic photograph of a movie scene, looking up at a colossal spaceship from a low position over a city --ar 16:9 --v 6.0 

下はstyle rawをonにした場合です。

s100 raw on

同様にon/offを順番に並べていきます。

off:

s50 raw off

on:

s50 raw on
Hyper-realistic photograph of a movie scene, looking up at a colossal spaceship from a low position over a city --ar 16:9 --v 6.0 --s 50 --style raw

off:

s150ではリアル (実写) ではないイラスト風のものが出始めました。

on:

s150 raw on

off:

s250 raw off

on:

s250 raw on

s500では逆にonでイラスト調 offで実写調となりました…

off:

s500 raw off (upscaled)

on:

s500 raw on

off:

s1000 raw off

on:

平均すると raw on実写調, raw offイラスト調のものができやすいと思われます。今回のシーン, 条件では優劣は分かりませんでした。

3.4 style rawとstylize値の設定について考察

両者二つをまとめると以下の通りになります。

① --style raw のon/off
on: プロンプトに忠実で写実的, 実写に近い。
off: 芸術性が強調される。
② --stylize値
: アートスタイルとエフェクトの多様性を強調。
: プロンプト忠実性と写実性を高める。

これだけ見れば写実性, リアリズムを高めるにはstyle raw をon,  stylizeを低めに設定となります。しかし, アートスタイルの多様性が損なわれるという風にあっちを立てればこっちが立たずと 頭が混乱してきました。分からないことはChatGPTに聞いてみる…と

SF映画の画像にふさわしい, style rawのon/off, stylize値の設定について。
リアリズム, 写実性, アートスタイル, 多様性の観点からまとめてください。

① (--style raw のon/off) と② (--stylize値) はどちらも芸術性と写実性に影響を与えますが ① (--style rawのon/off) はデフォルト加工の有無によって写実性に大きな影響を与えます。② (--stylize値) も影響を与えますが, 1に比べては微調整の範囲です。

写実性に関しては, --style raw onに加えて--stylize値を低く設定することも一つの方法ですが, 実際には--stylize値の影響は--style rawの設定による影響よりも限定的です。したがって, アートスタイルの多様性, 芸術性, および写実性の総合的バランスを取るためには--style raw on--stylize値高く設定する方が優れた結果をもたらす可能性があります。

ChatGPTの回答

SF映画の良質画像の定義は様々ですが, 個人的にはリアリズム, 実写に近く, アートスタイル, 多様性に富んだものが良いと思います。総合的にバランスをとるためには style rawon, stylize値は高めが良いとのことです。理由については分かったような分からないような…視覚的に表してもらいました。

横軸にリアリズム, 縦軸に多様性(アートスタイル, エフェクト, 創造性) の散布図。
style raw on/off, stylize 高低の4通りで視覚的にグラフで表してください。

ちょっとわかってきました。さらに次のようにも言っています。raw onとoffの違いがはっきり分かるのは stylize値が40から60の間とのことです。

実際実験してみても style rawon , stylize値100より大きめ (ChatGPTのいうstylize値高め) が『リアルで芸術性』が高そうなものが出力されることが多かったので本記事では基本的にstyle rawonをデフォルトとしたいと思います。

※完全に蛇足となりますが, 単純な発想としてプロンプトで弱点を補い制御できないか実験してみました。つまり, stylize値を高くして (1000), プロンプトリアルさを補う, または, Stylize値を低くして (0), プロンプトに芸術性を高める単語を含めて芸術性を補うというものです。

まずはStylize値低くして (0に設定), プロンプトに芸術性を高める単語 (アールヌーボー, 虹色) を入れてリアルさ芸術性同時に出すことができるか検証です。

プロンプト
(英) Hyper-relalistic, photorealistic photograph for an SF movie:A steampunk cityscape with a descending ornate spaceship merges warm glows and historical attire, as iridescent lights from an Art Nouveau vessel illuminate the cobblestones below --ar 16:9 --v 6.0 --style raw --s 0

(和) SF映画のためのハイパーリアリスティック, フォトリアリスティックな写真:華やかな宇宙船が降り立つスチームパンクな都市景観が, 暖かい光と歴史的な衣装を融合させ, アールヌーボー様式の船からの虹色の光が下の石畳を照らします

s0

実際出力された画像は芸術性の高い, イラスト調の画像で写真的なリアルさは全くなくなりました。これの理由としてChatGPTと一緒に出た結論が以下です。

① プロンプトに含めた『芸術性を高める』単語がイラスト的な画像を生成する傾向になること ② そもそもMidjourneyがそのような画像を事前学習していない, の2点です。個人的にはの要素が大きいと考えています。

次はstylize値を高くして (1000に設定), プロンプトにリアルさを高める方法です。下はstylize1000でリアルさを含まないプロンプトです。

Looking up at a colossal spaceship from a low position over a city --ar 16:9 --s 1000 --v 6.0 --style raw

s1000

実験のため『これでもか』というくらいのリアルさを高める以下のプロンプトを文頭に追加しました。

(英) Hyper-realistic, photorealistic image for an SF movie: Ultra high definition and lifelike accuracy render a futuristic cityscape. Vivid colors and unparalleled realism in 8K resolution deliver a crystal-clear, true-to-life representation with ultra-realistic textures.

(和) SF映画用のハイパーリアリスティック, フォトリアリスティックな画像:超高解像度と生命のような正確さで描かれる未来的な都市の景観。8K解像度での鮮やかな色彩と比類ないリアリズムが, 超リアルなテクスチャーを備えたクリスタルクリアで実物のような表現を実現します。

大分リアルさが向上しましたが, いわゆる『実写』ではなく, 現実に存在しそうな人々, リアルに存在しそうな建築物が追加されたような印象です。

s1000

理由は先ほどと全く同じことを繰り返しているだけなのですが, ① 芸術性を高める (Stylize値を高く設定) ことがMidjouneyの内部設定的にも, イラスト調にしリアルさを抑えることなること ② そもそも事前学習にそのような画像がない ということが挙げられます。

以上, リアルさ最重点に置いたMidjourney画像生成のコツは以下になります。

リアル画像を出力するコツ

1. style raw (rawモード)
:
style rawはonに設定して, デフォルトの加工を抑え, より自然で写実的なテクスチャとディテールを保持する。
2. stylize値
stylize値は中間に保ち, 過度なアートスタイルや抽象性を避ける。
3. 奇抜なシーンを避ける
AIの事前学習データに存在する要素を活用し, 過去に例のない奇抜なシーンはプロンプトから避ける

リアルな画像を出力するこつ

3.5 --no

画像に含めて欲しくないものを指示します (ネガティブプロンプト)。 item1とitem2を含めて欲しくないときは文末に『--no item1, item2』と追加します。下は実写の一人称の運転シーンとだけプロンプトしたものです。

種々のバリエーションのシーンが出てきました。ここで近未来感をだし, 落ち着いた印象にするためハンドル, 店, オレンジ色, ネオンライトを含めない指示をしてみます。

Hyper-realistic, photorealistic photograph for an SF movie: A first-person perspective inside a futuristic police vehicle --ar 16:9 --v 6.0 --style raw --no handle, orange, shops, neon light

プロンプトの文末に『--no handle, shops, orange, shops, neon light』を追加しました。

オレンジ色や派手なお店が除去され落ち着いた印象になりました。

4. プロンプト

4.1 基本の型

公式マニュアルによるとプロンプトの『構成要素』として『1. 主題 (Subject) 2. 媒体 (Medium) 3. 環境 (Environment) 4. 照明 (Lighting) 5. 色 (Color) 6. ムード (Mood) 7. 構図 (Composition)』が挙げられていました。

1. 主題 (Subject): 人物 (person), 動物 (animal), キャラクター (character), 場所 (location), オブジェクト (object)

2. 媒体 (Medium)
: 写真 (photo), 絵画 (painting), イラスト (illustration), 彫刻 (sculpture), 落書き (doodle), タペストリー (tapestry)

3. 環境 (Environment): 屋内 (indoors), 屋外 (outdoors), 月上 (on the moon), 水中 (underwater), 都市内 (in the city)

4. 照明 (Lighting): 柔らかい (soft), アンビエント (ambient), 曇り (overcast), ネオン (neon), スタジオライト (studio lights)

5. 色 (Color): 鮮やか (vibrant), 抑えられた (muted), 明るい (bright), モノクロ (monochromatic), カラフル (colorful), 白黒 (black and white), パステル (pastel)

6. ムード (Mood): 落ち着いた (sedate), 穏やか (calm), 騒がしい (raucous), エネルギッシュ (energetic)

7. 構図 (Composition): ポートレート (portrait), 顔写真 (headshot), クローズアップ (closeup), 鳥瞰図 (birds-eye view)

Midjourney公式 manualより

本記事ではプロンプトを主にChatGPTで作成していますが, 原案としてはこちらがまず大体のテーマを提示し, 専門的な描写をChatGPTに補完してもらっている形となっています。下は火星に降り立つ宇宙服の二人の映画のシーンとだけプロンプトしたものです。

Hyper-realistic, photorealistic photograph for an SF movie: Two astronauts standing on the surface of Mars, upper body close-up. --ar 16:9 --v 6.0 --style raw --s 250

1. 主題 (Subject) 宇宙飛行士のシーン, 2. 媒体 (Medium) リアルな映画, 3. 環境 (Environment) 火星は原案としてこちらが用意しましたが, 4.5.6ChatGPTに考えてもらいます。

明らかに映画 (プロモーション) らしくなっています。

Hyper-realistic, photorealistic photograph for an SF movie: Two astronauts standing on the surface of Mars, upper body close-up. The scene is bathed in the soft, diffuse light of the Martian day, casting gentle shadows and illuminating the astronauts' faces and suits. The color scheme is dominated by the rusty reds and oranges of the Martian soil, contrasted with the stark white and reflective materials of the space suits. The mood is contemplative and pioneering, capturing a moment of human achievement and the solitude of space exploration. The composition centers on the astronauts, focusing on their expressions of determination and wonder, with the Martian horizon stretching out behind them. --ar 16:9

4.2 リアルさを出すプロンプト

Promptにどのような単語を入れれば『写真らしさ, リアルさ』が出るかをChatGPTに聞いてみました。

SF映画用のシーンで写真のようにリアル, 高解像度な画像を再現するための単語
いままで出てきたプロンプトであなた (ChatGPT) が思う有用な単語を理由とともに順位づけしてください。

1. 超写実的 (Hyper-realistic) - 最もリアルなビジュアル表現を追求する基本的な要素。
2. 写真リアルな画像 (Photorealistic photograph) - 写真と見分けがつかないレベルのリアリズムを目指す。
3. 非常に高度な詳細 (Extremely highly detailed) - 細部までの精密さとディテールの豊かさを強調。
4. 4K - 高解像度で細部の鮮明さを保ちながらリアリズムを高める。
5. 高精細 (Ultra high definition) - 4Kに次いで、画像のクリアさとテクスチャの詳細を向上させる。
6. 生命感あふれる精度 (Lifelike accuracy) - リアルな質感、照明、影を通じて生命感を表現。
7. 高忠実度ビジュアル (High fidelity visuals) - 色彩の正確さとビジュアルの忠実度を高める。
8. クリスタルクリアな明瞭さ (Crystal-clear clarity) - 画像全体の明瞭さと視覚的鮮明さを確保。
9. 鮮やかな色彩 (Vivid coloration) - リアルな色彩表現で視覚的魅力を加える。

リアルさを出すプロンプト:GPTがつけた順位

今回の記事は『Hyper-realistic, photorealistic photograph (ハイパーリアルで写真のようなリアリズムを持つ画像)』を含めたプロンプトが多いです。Midjourneyは基本的に特に指示を与えなくても『リアル系』の画像を出力することが多いです。

サイバーパンクワールドの少女と指示

下の図は『Hyper-realistic, photorealistic photograph』を加えたものです。先ほどと比べ, リアルさが増しているのが分かります。

Hyper-realistic, photorealistic photograph:A girl in a cyberpunk world --style raw --v 6.0 --s 250

しかし, 『Hyper-realistic, photorealistic photograph』などのリアルさを出すプロンプトは常に必要ではなさそうです。以下はサイバーパンク以外にも『電気をまとったなど』いろいろ説明文を加えたものです。

リアルプロンプトなし
A cyberpunk girl with a helmet surrounded by electronics, a transparent panel shows her calm face. Her skin glows softly with digital light flickers. Her deep blue eyes are intense. Wires subtly frame her face, blending cool blues and grays for an enigmatic mood. Her expression is a mystery, enveloped in technology. --ar 16:9 --v 6.0 --s 250

『Hyper-realistic photo』というプロンプトを入れた場合と入れなかった場合で大差ありませんでした。

リアルプロンプトあり
Hyper-realistic photo: A cyberpunk girl with a helmet surrounded by electronics, a transparent panel shows her calm face. Her skin glows softly with digital light flickers. Her deep blue eyes are intense. Wires subtly frame her face, blending cool blues and grays for an enigmatic mood. Her expression is a mystery, enveloped in technology. --ar 16:9 --v 6.0 --s 250

理由についてChatGPTに聞いてみました。(※過去に大量にやりとりをしているChat欄のため筆者の意見が強めに反映されている可能性あります)

① Midjouneyが事前学習したデータには実写調の画像が多く, それが選ばれやすい。② プロンプトに実写調が選ばれる表現 (ビジュアル, 光, 感情の詳細) が多い 。とのことです。影響度を100点満点で表現してもらうと① 70点 ② 30点でした。

まとめですが

Hyper-realisticなどの『リアルさ』を強調する単語をプロンプトに含めると, 実写調の画像を生成しやすくなる。
プロンプトにビジュアル, 光, 感情の詳細を表す言葉が含まれている場合,『リアルさ』を特に強調しなくても実写調の画像が生成されやすくなる。
作成したい画像のスタイルが事前学習データにおいて主にイラスト調である場合, プロンプトでどれだけ努力しても, 実写調の画像を得ることは困難です。

(4.3 重み付け Weight)

要素を強調する機能です。基準値は『1』で通常のプロンプトは全ての単語に『重み1』が付加されているという考えです。

ChatGPT:重みづけの説明

例えば猫型ロボットを作成するとき, 要素どれくらい入れるかというときに比率で指示することができます。アンドロイド (ロボット) 1に対して猫を2倍重みをつけてみます。

cat::2 , android::1
Hyper-realistic, photorealistic photograph: A black and white cat-shaped android robot, cat::1 , android::0.5 --ar 16:9 --v 6.0 --style raw

アンドロイドの主張が強すぎますのでもっと減らしたいです。下は猫1 アンドロイド0.1です。

まだまだ強いですね。なかなかうまくいきません。一方, 下は猫0.1 アンドロイド1です。逆にしたら, 猫の要素が全くなくなりました。

cat::0.1 , android::1

使いこなせたら自由オリジナルキャラを作成したりと表現の幅が広がりそうですが, 今回検証した中では挙動が良く分からなかったので今回は使わないでおきます…

4.4 イメージプロンプト Image prompts

Midjourneyはかなりの画像再現能力があり, 画像ファイルをもとに正確なプロンプトがあれば『同じ』画像を作り出すことも可能となります。

手順としては, GPT4Vにファイルを画像アップロードし, 『この画像を再現するためのプロンプトを作成してください』と指示しプロンプトを用意します。Midjourneyのイメージプロンプトから画像アップロードしURLとGPTにつくってもらったプロンプトを渡します。

操作方法です。

Chat欄左の⊕マークをクリックし手持ちの画像ファイル『ファイルをアップロード』からMidjourneyに渡します。

先ほどの宇宙船を眺める女性を渡しました。ここで『Enter』を押します。次はこの画像URLを取得します。※URLは2種類あるので注意です。

間違えやすいのですがクリックから出てくる『メッセージリンク』ではありません。

一回, 画像を左クリックし画像ウィンドウがでてから右クリック『画像リンクをコピー』からURL取得。

その他にも取得方法はいくつかあり, 画像をクリックしブラウザで開き

ブラウザの上に出てくるURL欄からコピーして取得。

ブラウザを開いた状態でクリック『画像リンクをコピー』から取得するなど。(公式マニュアルにも他の方法でうまくいかなかったらこの方法が推奨されています。)

<https://s.mj.run/ft_■■■■> Hyper-realistic, photorealistic photograph for an SF movie: A futuristic woman stands on the high balcony of a building, painted in shades of dark gray and dark bronze, overlooking a starship and a futuristic cityscape. This scene carries the ambiance of a movie still, with clusters of rugged buildings, intricate military details, translucent flying vehicles, and the anticipation of imminent intense action scenes. --ar 16:9 --v 6.0 --style raw

『/imagine』コマンドから, まずURLを貼り付け (ピンク色), 続いて状況説明のプロンプト, パラメーター (青) を入力していきます。

ここで注意なのは画像だけではうまくいかず, しっかりと状況を説明するテキストプロンプトが必要なことです。さきほどChatGPTに作ってもらったプロンプトを流用しました (2.3 /describe参照)。

見事に再現されています。

しかし元画像自体がMidjourney製なので, 出来て当たり前のような気がしてきましたので, あまり知られていない映画奇抜なシーンではどのようになるのか調べてみました。

『エンジェルウォーズ』という映画 (©Warner Bros Pictures) の予告編からです。

同様にChatGPTに状況説明のプロンプトを書いてもらい。MidjourneyのImage Promptsでこの画像を渡しただけです。プロンプトに関しては他の画像作成時と同様のtemplateを使っています (6章参照)。

<https://s.mj.run/VWF8■■> Hyper-realistic, photorealistic photograph for dark fantasy SF movie: A woman dressed in a navy blue sailor suit stands atop a futuristic, devastated battlefield, a sword strapped to her back. The camera captures her from a low position behind, highlighting her upper body against a backdrop of desolate terrain. The moody dark sky and the monochromatic blue-gray tones emphasize the bleak atmosphere. --ar 16:9 --v 6.0 --style raw

4.5 カメラの設定 

ChatGPTSF映画相性の良いカメラ設定を聞いてみました。広い絞り (Wide Aperture), ボケ効果 (Bokeh Effect), HDR (High Dynamic Range), 中間密度フィルター (ND Filter) がおすすめとのことです。 

8番目のカテゴリーにカメラ設定も加えてみます。

8. カメラ設定 (Camera Settings): 広い絞り (Wide Aperture), ボケ効果 (Bokeh Effect), HDR (High Dynamic Range), 中間密度フィルター (ND Filter)

1. 広い絞り(Wide Aperture): 被写体を際立たせ, 背景ぼかして強いボケ効果を生み出します。
2. ボケ効果(Bokeh Effect): 背景のぼかしを通じて, 視覚的な焦点被写体に集中させ, 物語の重要な要素を強調します。
3. HDR(High Dynamic Range): 明るい部分と暗い部分の詳細を保持し, リアルな画質でシーンのダイナミクスを向上させます。
4. 中間密度フィルター(ND Filter): 強い光の下でも露出をコントロールし, 絞りを広げてボケ効果を得ることができます。

カメラ設定の種類

基本となる画像を作り, 様々なカメラ設定のプロンプトに加えてみます。

基本画像 s250

1. 広い絞り  wide aperture
sharply in focus due to the use of a wide aperture setting (広い絞り設定の使用により鮮明に焦点が合っている) を加えています。

Hyper-realistic, photorealistic photograph for a fantasy SF movie: An elf stands on an expansive grassland, sharply in focus due to the use of a wide aperture setting. This creates a beautiful bokeh effect, slightly blurring the castle town in the distance. The medieval architecture of the town, with hints of futuristic technology, melds into the soft background, emphasizing the elf's connection to both the natural and the technologically advanced worlds. The lighting and color palette bring out the vibrant greens of the grass and the subtle contrasts between the elf's attire and the environment. --ar 16:9 --v 6.0 --style raw --s 250

エルフ焦点が合い, 遠くの城下町をぼかしています。

2. ボケ効果 Bokeh effect

ChatGPTにボケのカメラ効果を加えた文を英訳してもらいます。

背景をぼかして円形, 多角形のも表現しているとのことです。

3. HDR(High Dynamic Range)

画像のコントラストが高まり, 草木1本1本の質感がはっきりと分かる, また, 遠くの城のの部分も失われずはっきりと描写されるとのことです。

4. 中間密度フィルター(ND Filter)

さらに詳細設定したい人はこちらの記事がおすすめです。

5. Midjourneyの得意, 不得意

リアル写真調の画像生成という点でどのようなシーンが出来て, どのようなシーンが出来ないかの考察となります。

夜のドライビングシーンです。描写は苦手で実写系の夜の運転の描写はプロンプト, パラメーターをいじってもなかなかうまくできませんでした。

下はプロンプトから『夜の at night』を抜いたものです。アップスケール (高解像度) 処置はしていますが明らかにイラスト調とは異なります。

Upscaled
Hyper-realistic, photorealistic photograph for an SF movie, taken from a first-person perspective within the driver's seat. The driver is partially visible, adding a personal touch to the futuristic urban landscape. --ar 16:9 --v 6.0 --style raw --s 250

その他にもファンタジー色, 芸術色が強いものはシーンはそのままで『イラスト調』で再現されたり, 河童など和風のキャラクターは実写を維持しつつもべつの『西洋の映画に出てくるトカゲ』に置き換えられたりします。

これらの苦手な分野は事前には分からず, 実際にプロンプトしてみてどのように出力されるか個別に実験する必要がありそうです。

6. 制作例 プロンプト例

基本のテンプレートとしては, 簡単を出し, ChatGPTに英訳してもらっているだけです。書き出しを『Hyper-relalistic, photorealistic photograph for an SF movie:』『SF映画用の超写実的で写真のようにリアルな写真』としています。

原案をこちらが用意しますので短めの英文プロンプトを作成してください。■には状況説明が入ります。
例) Hyper-realistic, photorealistic photograph for an SF movie: ■ --ar 16:9

望み通りの画像が得られることもありますが 基本的にはスタイル, 色, ムードを追加するように言います。

4. 照明 (Lighting) 5. 色 (Color) 6. ムード (Mood) のプロンプトも追加してください。

それでもうまくいかない場合は, これらの設定を個別に設定します (後述)。

スタイライズ値はmid (100), high (250) を基本としあまりにも写真寄りで質素な場合は, 効果, 多様性をつけるために250より大きい値を採用しました。

Stylize mid (デフォルト)          100
Stylize high                               250

原則 --style raw 有効としています。

再現性が重要と考えていますので, まぐれで出てきた画像は掲載していません。1-2回で出てきた画像のみを採用しています。

宇宙人襲来

某SF映画リスペクトです。巨大な宇宙船がホワイトハウスを破壊するシーンとなります。まずはシンプルプロンプト。

s250

4. 照明 (Lighting) 5. 色 (Color) 6. ムード (Mood) をプロンプトに含めてもらいます。同じchat欄で会話していますので『456指定』だけで伝わります。

s300
Hyper-realistic, photorealistic photograph for an SF movie: A close-up scene where a gigantic spaceship, occupying the upper half of the frame, is destroying the White House, situated in the lower half, by firing a blue beam directly downward. The stark contrast between the cool, destructive blue of the beam and the traditional whites of the White House emphasizes the scene's tension. The mood captures the shock and urgency of the moment, highlighting the unexpected disaster. --ar 16:9 --v 6.0 --style raw --s 300
Hyper-realistic, photorealistic photograph for an SF movie: A futuristic woman on a high-rise gazes at a starship over a city that blends dark gray and bronze hues. The composition merges raw urban textures with the sleekness of advanced technology, setting a scene ripe with anticipation and detailed military elements. --ar 16:9 --v 6.0 --style raw --s 250

下の図は以前作成した画像のプロンプトを忘れたたImage Promptsで再現するようにお願いしたものです。

<https://s.mj.run/OhHjJOHiJms> Hyper-realistic, photorealistic photograph for an SF movie: A futuristic woman stands on the high balcony of a building, painted in shades of dark gray and dark bronze, overlooking a starship and a futuristic cityscape. This scene carries the ambiance of a movie still, with clusters of rugged buildings, intricate military details, translucent flying vehicles, and the anticipation of imminent intense action scenes. --ar 16:9 --v 6.0 --style raw --s 250

未来の警察もの

大会議室みたいなところで会議していてイメージが違いましたので

『少人数グループ small group』『上半身アップ upper body close up』に変更しました。

s250
Hyper-realistic, photorealistic photograph for an SF movie: A small group of officers engage in a meeting within a futuristic police station, focused on a holographic display. The scene is captured in an upper body close-up, highlighting the interaction with advanced hologram technology in the intimate meeting space. --ar 16:9 --v 6.0 --style raw --s 250

2024年宇宙の旅

宇宙船内部の宇宙飛行士

浮かんでくれませんでした。

s750
Hyper-realistic, photorealistic photograph for an SF movie: Inside a futuristic spaceship, astronauts float in zero gravity, effortlessly moving through the cabin. --ar 16:9 --v 6.0 --style raw --s 750

宇宙ステーションから見た地球

宇宙ステーションが小さく臨場感がありませんでした。『画面いっぱいに fills the frame』と追加しました。

s250
Hyper-realistic, photorealistic photograph for an SF movie: The International Space Station fills the frame, with Earth's stunning view as the backdrop, highlighting the planet's curvature and vibrant colors from orbit. --ar 16:9 --v 6.0 --style raw --s 250

火星に降り立つ二人

こちらは『火星に降り立つ二人の宇宙飛行士』とだけChatGPTに与え, 4. 照明 (Lighting) 5. 色 (Color) 6. ムード (Mood) 残りのプロンプトをお任せしたものです。

Hyper-realistic, photorealistic photograph for an SF movie: Two astronauts standing on the surface of Mars, upper body close-up. The scene is bathed in the soft, diffuse light of the Martian day, casting gentle shadows and illuminating the astronauts' faces and suits. The color scheme is dominated by the rusty reds and oranges of the Martian soil, contrasted with the stark white and reflective materials of the space suits. The mood is contemplative and pioneering, capturing a moment of human achievement and the solitude of space exploration. The composition centers on the astronauts, focusing on their expressions of determination and wonder, with the Martian horizon stretching out behind them. --ar 16:9 --v 6.0 --style raw --s 250

火星の砂埃で視界がぼやける

以下のサイトを参考にしています。

まずは『火星で男が一人砂埃で視界がややぼやける』とだけプロンプト。

4. 照明 (Lighting) 5. 色 (Color) 6. ムード (Mood) を追加したプロンプト

Hyper-realistic, photorealistic photograph for an SF movie: A lone astronaut on Mars, with the vision subtly blurred by a faint haze of sand, not so much a sandstorm but enough to make distant views hazy. The lighting is soft and diffused, gently illuminating the landscape and the astronaut. Colors are desaturated, with pale reds and dusty oranges of the Martian ground, and a subtle contrast from the astronaut's suit. The mood is serene yet somber, emphasizing solitude in the vast, quiet expanse of Mars --ar 16:9 --v 6.0 --style raw --s 250

砂埃, 全体的なスタイルはイメージに近づいてきましたがオレンジ色が欲しいところです。

Hyper-realistic, photorealistic photograph for an SF movie: A lone astronaut on Mars, with vision subtly blurred by a faint haze of sand, not so much a sandstorm but enough to soften distant views. The lighting is soft and diffused, casting a warm glow across the Martian landscape and astronaut. Colors are rich in shades of orange, enhancing the dusty reds of the Martian ground with a more pronounced orange hue, creating a vibrant contrast against the astronaut's suit. The mood is tranquil yet poignant, highlighting the astronaut's isolation amidst the vast, silent expanse of Mars, bathed in an orange-tinted light --ar 16:9 --v 6.0 --style raw --s 250

ファンタジーSF

登場人物を作っていきます。

 4. 照明 (Lighting) 5. 色 (Color) 6. ムード (Mood) 追加します。

Hyper-realistic, photorealistic photograph for an SF movie: A young knight in a SF fantasy setting, his stance exuding readiness for adventure. The lighting casts dramatic shadows, emphasizing the fusion of his traditional armor with futuristic enhancements. The color palette combines the metallic sheen of his gear with the vibrant energy of a world where fantasy and science fiction intersect. The mood is charged with anticipation and valor, capturing a moment where past legends and future technology collide. The composition frames the knight against a backdrop that blurs the lines between the ancient and the cosmic, inviting viewers into a realm of endless possibilities. --ar 16:9 --v 6.0 --style raw --s 250

同様にエルフの魔法使いです。

実写系のアップシーンはカメラ設定も有効とのことです。以下の記事を参考にしました。

『魚眼レンズ Captured with a fish-eye lens for a dynamic effect』の設定を加えています。

Hyper-realistic, photorealistic photograph for an SF movie: An upper body shot of a female elf, tailored for a classic fantasy theme. The lighting is warm and balanced, casting a natural glow that highlights her intricate elven features and the detailed craftsmanship of her attire. The color scheme embraces the rich, earthy tones of the forest, combined with the subtle shimmer of magic inherent to her being. The mood is one of adventure and mystique, capturing the timeless appeal of fantasy lore where elves and magic play a central role. The composition carefully balances her against a backdrop that suggests an ancient, enchanted woodland, inviting viewers into a world of wonder and exploration. Captured with a fish-eye lens for a dynamic effect. --ar 16:9 --v 6.0 --style raw --s 250

中世ファンタジーSF 若い騎士 エルフ ドワーフが酒場で会議とプロンプトです。

カメラを少し引き気味 『A slightly zoomed-out view of』と追加しました。

Hyper-realistic, photorealistic photograph for a fantasy SF movie set in the medieval era: A slightly zoomed-out view of a medieval tavern where a young knight, a female elf, and a dwarf are seen strategizing around a large wooden table. The tavern combines elements of medieval architecture with discreet futuristic technology, offering a scene where past and future coexist harmoniously. The wider angle captures the bustling atmosphere of the tavern, with other patrons in the background hinting at a world alive with adventure and intrigue. The group at the table is lit by a combination of candlelight and the soft, otherworldly glow from a high-tech device, highlighting the fusion of ancient traditions and advanced civilizations. This photograph encapsulates a moment of unity and purpose among diverse beings in a world where magic and technology intertwine. --ar 16:9 --v 6.0 --style raw --s 250

草原 エルフ 城下町とのみ指示したものです。

Hyper-realistic, photorealistic photograph for a fantasy SF movie: An elf stands on an expansive grassland, gazing towards a castle town in the distance. The town combines medieval architecture with subtle hints of futuristic technology, bridging the worlds of fantasy and science fiction. The vast grassland foregrounds the elf, highlighting their connection to both the natural and the crafted worlds. --ar 16:9 --v 6.0 --style raw --s 250

洞窟 骸骨の化け物と戦闘シーンです。
SF映画というよりはファンタジー3D ゲームの映像に近いです。

Hyper-realistic, photorealistic photograph for a fantasy SF movie: In a shadowy cave, a knight and an elf wizard confront a skeleton monster. The lighting, a mix of the wizard's spell and the natural dimness of the cave, highlights the intense expressions and dynamic poses of the characters, casting long shadows that add depth to the scene. The color palette is dominated by cool blues and eerie greens from the magical glow, contrasting with the dark, earthy tones of the cave and the stark white of the skeleton. The mood is tense and suspenseful, capturing the moment of clash between good and evil. The composition skillfully focuses on the trio, with the knight and wizard in the foreground, symbolizing their courage and unity against the looming threat in the background. --ar 16:9 --v 6.0 --style raw --s 250

5. 得意不得意で考察したように事前学習で似たような映画がないため, ゲームのワンシーンを再現したような画像が生成されていると思われます。そのため実写よりはゲーム用の3Dグラフィックにテイストが似ています。

Hyper-realistic, photorealistic photograph for an SF movie: A hero in combat with a demon lord in a futuristic battleground. The scene is a dramatic clash of good versus evil, set in an environment that blends elements of advanced technology with mystical forces. The hero, equipped with high-tech armor and weapons, faces off against the demon lord, a formidable foe with dark powers and a menacing appearance. The battleground is filled with the remnants of ancient ruins and futuristic machinery, highlighting the timeless battle between light and darkness. This encounter captures the essence of a sci-fi fantasy epic, where the fate of the world hangs in the balance --ar 16:9 --v 6.0 --style raw --s 250

アンドロイド人工知能

ChatGPTに『サイバーパンク』など適当に単語を渡し作ってもらいました。

A girl in a cyberpunk world her helmet surrounded by advanced electronics with a transparent panel revealing a composed face Her skin emits a soft glow dotted with gentle flickers of digital light Eyes deep blue compelling and intense Wires contour her helmet subtly framing her features The color scheme blends cool blues and grays setting an enigmatic mood Her expression remains inscrutable a riddle wrapped in technology's embrace. --ar 16:9 --v 6.0 --style raw --s 250

よく見かける ピンクをベースにしたサイバーパンク風SFです。

A robotic entity stands amidst the rain in a neon-lit alley, encapsulating the essence of a lively Asian city at night. Its design is humanoid, showcasing a sleek combination of black and striking pink, mirroring the wet, shimmering surroundings. The robot's head features elaborate designs, glowing with a soft pink hue. Neon signs, adorned with Asian characters, illuminate the scene, enveloping it in a radiant fog under the night sky. The alley, unusually empty, contrasts sharply with the expected vibrancy of such an urban setting. Raindrops enhance the scene's reflective beauty, casting a symphony of light and shadow that accentuates the robot's polished appearance --ar 16:9 --v 6.0 --style raw --s 250

ロボットのアップ

A cybernetic figure stands prominently in a rain-drenched, neon-lit street, reminiscent of a bustling Asian metropolis. The robot is sleek and humanoid, with a high-gloss finish in shades of black and vibrant pink that reflect the wet urban landscape. Its head displays intricate details and is illuminated by a radiant pink glow, with eyes that seem to be gazing into the distance. Neon signs with Asian characters hang in the background, contributing to the atmosphere of a cyberpunk cityscape. The street is empty, suggesting a moment of solitude in the otherwise vibrant city life. The rain adds a reflective quality to the surfaces around, enhancing the vivid colors and the contrast between the organic and the mechanical. --ar 16:9 --v 6.0 --style raw --s 250

ベースです。

Hyper-realistic, photorealistic photograph for an SF movie: A cybernetic woman in profile, her silhouette sharply outlined against a console illuminated in vibrant shades of blue. The console, alive with digital readouts and interfaces, showcases advanced technology at its core. She seems to be an integral part of this system, with her body displaying a seamless fusion of organic elegance and cybernetic enhancements that echo the complex circuitry patterns behind her. Set within an industrial, yet utilitarian environment, her presence introduces a sense of purpose and grace. The atmosphere buzzes with the energy of data streams, while the cool blue lighting casts a futuristic glow, emphasizing the symbiosis of human and machine --ar 16:9 --v 6.0 --style raw --s 250
A youthful android with a visage echoing that of a young woman occupies the center of a serene scene, her eyes alight with a consciousness that belies her synthetic origins The suit she wears is devoid of color in this description, allowing for a myriad of potential shades that would harmonize with the tranquil blue neon emanating softly from behind Her features are poised between the natural and the artificial, against a backdrop that whispers of advanced technology the quiet hum of a spaceship or the hushed operations of an orbital platform --ar 16:9 --v 6.0 --style raw --s 250

電子回路のデザインの服を着たアンドロイド

A woman stands in a futuristic setting, her upper body elegantly fused with a cybernetic structure that features electronic circuitry in cool blues and emerald greens. The advanced circuit design signifies high technology. Lighting is subtle and comes from a cool blue spectrum, softly illuminating her and reflecting off her metallic integration. The color scheme is composed of cool blues and greens, with strategic highlights accentuating the serene ambiance. The mood is tranquil and reflective, with the seamless blend of human and machine elements suggesting a peaceful coexistence within a modern, sophisticated tableau --ar 16:9 --v 6.0 --style raw --s 250

顔以外とプロンプト

A woman stands in a futuristic setting, her upper body, excluding her face, adorned with a cybernetic structure that integrates electronic circuit patterns in shades of blue and emerald green. The sophisticated circuitry highlights the advanced technology within her. The ambiance is serene, lit by cool blues that underscore the seamless integration of human and technological elements --ar 16:9 --v 6.0 --style raw --s 250

ロボット戦争もの

Hyper-realistic, photorealistic photograph for an SF movie: A futuristic spaceship adorned with ornate designs and a polished finish glides through a tempestuous, cloud-filled sky. The vessel's surface is intricately detailed, with dark silver and luminous emerald highlights that catch the storm's erratic lighting. The image captures the essence of motion and sophistication in an ultra-high-definition snapshot of power and elegance amidst the chaos of a cosmic maelstrom. --ar 16:9 --v 6.0 --s 250 --style raw

ChatGPTに『色 (Color) 光 (Lighting)の設定を詳細に加えてください。』とお願いしました。

Hyper-realistic, photorealistic photograph for an SF movie: A sleek spaceship maneuvers through a storm-swept sky, its hull reflecting the turbulent gray clouds and flashes of lightning. The craft is detailed with a palette of dark silvers that gleam with a metallic sheen, and emerald lights that punctuate its form with bursts of color. Ambient lighting from the storm casts dramatic shadows and highlights across the vessel, creating a dynamic interplay of light and color that emphasizes the ship's aerodynamic design. The intense chiaroscuro effect dramatizes the scene, underscoring the ship's journey through the heart of a galactic tempest. --ar 16:9 --v 6.0 --s 250 --style raw

砂漠でのロボット同士の戦闘シーンです。

Hyper-realistic, photorealistic photograph for an SF movie: Mechs resembling bipedal tanks advance through a desert, kicking up clouds of dust as they exchange heavy gunfire. The intense midday sun casts sharp shadows on the dunes, while explosions and muzzle flashes provide stark, contrasting light. A palette of desert browns and the dull grays of the mechs' armor is punctuated by the fiery oranges of weapons discharges, creating a scene fraught with tension and the chaos of battle. --ar 16:9 --s 250 --v 6.0 --style raw

コクピット内でロボット操縦シーンです。

Hyper-realistic, photorealistic photograph for an SF movie: The image pulls back to show a pilot within the vast cockpit of a towering combat robot. The pilot is encased in a halo of high-tech displays and controls, while the robot's armored exterior is visible through the panoramic cockpit glass, hinting at the scale and power of the machine. The scene is lit with a combination of interior ambient light and the harsher external light of battle, casting a diverse array of shadows and highlights across the cockpit's interior and the robot's outer shell. The color palette skillfully balances the cool interior lighting with the warmer, more dynamic tones of the battlefield, creating an atmosphere thick with tension and anticipation. --ar 16:9 --v 6.0 --s 250 --style raw

この記事が参加している募集

AIとやってみた