私的AI研究会 > ComfyUI14
「ComfyUI」を使ってローカル環境でのAI画像生成を検証する
| 2026年5月 中国のAIスタートアップ「HiDream.ai」からオープンソースとして正式に発表・公開されされた画像生成モデル「HiDream-O1-Image」を検証してみる |
| モデル名 | 特徴 |
| HiDream-O1-Image full | 標準モデル 50 steps, guidance scale 5.0, shift 3.0 |
| HiDream-O1-Image dev | 蒸留モデル（dev）は 28 steps, guidance scale 0, shift 1.0 |
| Width | Height | Category |
| 2048 | 2048 | Square |
| 2304 | 1728 | Landscape |
| 2304 | 1792 | Landscape |
| 2496 | 1664 | Landscape |
| 2560 | 1440 | Landscape |
| 3104 | 1312 | Landscape |
| 1728 | 2304 | Portrait |
| 1792 | 2304 | Portrait |
| 1664 | 2496 | Portrait |
| 1440 | 2560 | Portrait |
| 1312 | 3104 | Portrait |
| ・ HiDream-O1-Image は、Qwen3-VL という 大規模言語モデルを拡張して 画像生成能力 を追加したモデル ・ 言葉と絵を同じ次元で理解して（UiT：Pixel-level Unified Transformer アーキテクチャ）、画像を生成する ・ 潜在空間は存在しないので生成過程を逐次確認 することができ、復号の必要がなく復号に伴う画質の劣化もない |
| 「AI が効率よく画像を生成するには、潜在空間が必要」というのが、画像生成 AI が公開されて以来 ４年間の常識だったが、HiDream-O1-Image が UiT アーキテクチャのもとに、前作の ３分の１の容量のモデル でこの常識を打ち破ってみせた |
| HiDream-O1-Image は、言語と画像を同じ次元で処理 することにより、今までより 画像編集を自然にこなす ことができる |
| AI は、大量の情報を流し込むことによって人間が望む機能を習得sするが、その詳細は ブラックボックスだが、言語と画像を同じ次元で処理して、潜在空間を使わないで人間から処理が見える HiDream-O1-Image モデルが登場したことにより、これから 画像生成 AI モデルそのものの解明 が進むことも期待できるかもしれない |
| HiDream-O1-Image は、Alibaba の Qwen3-VL モデルをベースに作られている Qwen シリーズは、その性能の高さとオープンなライセンスから、2026年5月現在、画像生成のテキストエンコーダーとしての標準 の地位を確立しつつある |
| HiDream-O1-Image モデルは、すでに パラメータ数を 200B 以上に拡張 した 高性能モデル が存在することも明らかになっている。HiDream-O1-Image の MIT ライセンス で公開された UiT アーキテクチャは、これからの画像生成の標準になり、現行の画像生成 AI は UiT アーキテクチャのもとに再編 されるかもしれない |
| このプロジェクトで作成するワークフローと関連データは下記にアップロードしている（更新されている場合は再度ダウンロードのこと） |
📂ComfyUI
├─📂input ← ワークフローに含まれる入力画像
└─📂user
└─📂default
└─📂workflows ← ワークフローの保存場所
:
├─📂~beta
├─📂HiDream-O1 ← この章で作成するワークフロー
:
・解凍してできる「ComfyUI/」フォルダを「StabilityMatrix/Data/Packages/ComfyUI」へ上書きコピーする| ワークフロー | 機 能 | モデル | CPU | CPU | |||||
| RTX 4070 | RTX 4060 | RTX 4060L | RTX 3050 | GTX 1050 | i7-1260P | i7-1185G7 | |||
| 7001_hidream_o1_full | full(通常版)基本ワークフロー | full | 11:55.78 | 16:11.64 | 84:05.96 | ||||
| 7002_hidream_o1_full_simple | full(通常版)Simple ワークフロー | full | 03:50.20 | 05:06.22 | 19:34.46 | ||||
| 7010_hidream_o1_full_t2i | full(通常版) Text to Image | full | 03:51.36 | 12:24.36 | |||||
| 7011_hidream_o1_full_iti | full(通常版) Image to Image | full | 10:32.35 | 122:20.21 | |||||
| 7101_hidream_o1_dev | dev(蒸留版)基本ワークフロー | dev | 06:16.78 | 08:03.71 | 18:34.02 | ||||
| 7102_hidream_o1_dev_simple | dev(蒸留版)Simple ワークフロー | dev | 02:18.97 | 03:11.63 | 05:39.46 | ||||
| 7110_hidream_o1_dev_t2i | dev(蒸留版) Text to Image | dev | 01:19.16 | 02:45.37 | |||||
| 7111_hidream_o1_dev_iti | dev(蒸留版) Image to Image | dev | 02:43.84 | 06:38.05 | |||||
| モデル名 | ファイル名（.safetensors） | 配置先（/StabilityMatrix/Data/） | ダウンロード URL | |
| checkpoints | hidream_o1_image_fp8_scaled | Models/ | StableDiffusion/ | hidream_o1_image_fp8_scaled.safetensors |
| hidream_o1_image_dev_fp8_scaled | hidream_o1_image_dev_fp8_scaled.safetensors | |||
| loras | Lora/ | |||
| text_encoders | gemma4_e4b_it_fp8_scaled | text_encoders | gemma4_e4b_it_fp8_scaled.safetensors | |
| テンプレート名 | モデル | ワークフロー | 保存ワークフロー名 | |
| ④ | HiDream O1 Full Image... | full 通常版 | image_hidream_o1.json | image_hidream_o1_org.json |
| ⑤ | HiDream O1 Dev | dev 蒸留版 | image_hidream_o1_dev.json | image_hidream_o1_dev_org.json |
| ・テンプレートのオリジナル・ワークフローに対して「① モデルの変更, ② 入力画像の指定, ③ シード値を固定（同じ画像を再現できるように）」を変更する ・full 通常版 / dev 蒸留版 の2種類のテンプレートが用意されている。対応モデルは異なるが機能は同一 ・参照画像の 有効/無効 をノード「Switch to Image Edit」で切り替える（false の場合でもダミー画像は必要） ・プロンプト・エンハンサーの 有効/無効 をノード「Enable Prompt Refine?」で設定することができる（有効とした場合は英語のプロンプトとなる） |
 image_hidream_o1_org.json |  SubGraph: プロンプト・エンハンサー |
| （参考）プロンプト・エンハンサーにより内部で生成されたプロンプト | |
| A hyper-detailed, photorealistic close-up portrait of a graceful Japanese woman showcasing sophisticated skincare. She has flawless, dewy skin with light nude makeup, emphasizing natural beauty and hydration. She is delicately holding an elegant, minimalist glass essence bottle, positioned slightly off-center in her hands. The background is a seamless, warm ivory color, providing a soft, luxurious canvas that complements her complexion. The lighting is exceptionally soft and diffused, mimicking early morning sunlight filtering through sheer fabric, creating gentle highlights on her skin and the glass bottle. The composition utilizes a shallow depth of field (bokeh effect) to keep the focus razor-sharp on her face and the product, while the background melts into a creamy blur. Shot with an 85mm prime lens, high resolution, editorial beauty photography style. | 洗練されたスキンケアを披露する、優雅な日本人女性を写した、超高精細でフォトリアリスティックなクローズアップポートレート。彼女は、ナチュラルな美しさと潤いを強調する、軽やかなヌードメイクを施した、完璧でみずみずしい肌をしています。彼女は、エレガントでミニマルなガラス製のエッセンスボトルを、やや中心からずれた位置にそっと持っています。背景は継ぎ目のない温かみのあるアイボリーカラーで、彼女の肌の色を引き立てる、柔らかく贅沢なキャンバスとなっています。照明は非常に柔らかく拡散しており、薄手の布を通して差し込む早朝の太陽光を模倣し、彼女の肌とガラスボトルに柔らかなハイライトを生み出しています。構図は、浅い被写界深度（ボケ効果）を利用して、彼女の顔と製品にピントを合わせ、背景はクリーミーなぼかしで溶け込んでいます。85mm単焦点レンズを使用し、高解像度で、エディトリアルビューティーフォトグラフィースタイルで撮影されました。 |
 image_hidream_o1_dev_org.json | SubGraph: プロンプト・エンハンサー |
| （参考）プロンプト・エンハンサーにより内部で生成されたプロンプト | |
| A highly cinematic, photorealistic depiction of a dense, futuristic cityscape street scene at deep night. The entire background is transformed into a vibrant, rain-soaked neon metropolis. Heavy, continuous rain streaks down the frame, creating slick, highly reflective puddles on the wet asphalt ground. Blurry, saturated neon signs—emitting vivid cyan, magenta, and electric blue light—are reflected dramatically across the glossy, dark pavement, creating long, distorted streaks of color. The atmosphere is moody, cyberpunk, and intensely urban. Use a shallow depth of field to keep the immediate foreground slightly sharper while allowing the distant city lights to dissolve into beautiful bokeh spheres. Shot with an 85mm prime lens, low-key lighting, high contrast, and dramatic volumetric fog to enhance the feeling of damp, electric energy. | 真夜中の、未来的な都市の街並みを、映画のような写実的な描写で捉えた作品。背景全体が、雨に濡れた鮮やかなネオンの都会へと変貌する。激しく降り注ぐ雨が画面を流れ落ち、濡れたアスファルトの地面には、滑らかで反射率の高い水たまりができている。鮮やかなシアン、マゼンタ、エレクトリックブルーの光を放つ、ぼやけた彩度の高いネオンサインが、光沢のある暗い舗装路に劇的に反射し、長く歪んだ色の筋を作り出す。雰囲気は陰鬱で、サイバーパンク的で、都会的な雰囲気に満ちている。浅い被写界深度を用いることで、手前の被写体はややシャープに写し出しつつ、遠くの街の灯りは美しいボケ球へと溶け込む。85mm単焦点レンズ、ローキー照明、高コントラスト、そして湿った電気的なエネルギー感を高めるためのドラマチックなボリュームフォグを用いて撮影した。 |
| 7001 基本ワークフロー | 7001 SubGraph |
 |  |
| 「HiDream-O1/」7001_hidream_o1_full.json | |
| Graceful japanease female skincare shot, light nude makeup, holding essence bottle, warm ivory backdrop, soft diffused light | Transform the background into a rainy neon city street at night, with wet asphalt reflections, blurred neon signs | ||
| 優雅な日本人女性のスキンケアショット。薄めのヌードメイク、美容液ボトルを持ち、温かみのあるアイボリーの背景、柔らかな拡散光。 | 背景を雨に濡れた夜のネオン街に変え、濡れたアスファルトの反射やぼやけたネオンサインを描きます。 | ||
| Switch to Image Edit = false | Switch to Image Edit = true | ||
| Enable Prompt Refine? = true | Enable Prompt Refine? = false | Enable Prompt Refine? = true | Enable Prompt Refine? = false |
 |  |  |  |
| 画像サイズ：2048x2048 | |||
| ・プロンプト・エンハンサーは速度が遅くなるので削除する。必要であれば外部の LLM などで詳細なプロンプトを作成し入力とすることで補完できる ・検証の結果特に大きな問題はないと思われるので日本語のプロンプトを採用する |
| 7002 | 生成画像 | |
 | 優雅な日本人女性のスキンケアショット。薄めのヌードメイク、美容液ボトルを持ち、温かみのあるアイボリーの背景、柔らかな拡散光。 | 背景を雨に濡れた夜のネオン街に変え、濡れたアスファルトの反射やぼやけたネオンサインを描きます。 |
| Switch to Image Edit = false | Switch to Image Edit = true | |
| |  | |
| 「HiDream-O1/」7002_hidream_o1_full_simple.json | 画像サイズ：2048x2048 | |
| 7101 基本ワークフロー | 7101 SubGraph |
 |  |
| 「HiDream-O1/」7101_hidream_o1_dev.json | |
| Graceful japanease female skincare shot, light nude makeup, holding essence bottle, warm ivory backdrop, soft diffused light | Transform the background into a rainy neon city street at night, with wet asphalt reflections, blurred neon signs | ||
| 優雅な日本人女性のスキンケアショット。薄めのヌードメイク、美容液ボトルを持ち、温かみのあるアイボリーの背景、柔らかな拡散光。 | 背景を雨に濡れた夜のネオン街に変え、濡れたアスファルトの反射やぼやけたネオンサインを描きます。 | ||
| Switch to Image Edit = false | Switch to Image Edit = true | ||
| Enable Prompt Refine? = true | Enable Prompt Refine? = false | Enable Prompt Refine? = true | Enable Prompt Refine? = false |
 |  |  |  |
| 画像サイズ：2048x2048 | |||
| ・プロンプト・エンハンサーは速度が遅くなるので削除する。必要であれば外部の LLM などで詳細なプロンプトを作成し入力とすることで補完できる ・検証の結果特に大きな問題はないと思われるので日本語のプロンプトを採用する |
| 7102 | 生成画像 | |
 | 優雅な日本人女性のスキンケアショット。薄めのヌードメイク、美容液ボトルを持ち、温かみのあるアイボリーの背景、柔らかな拡散光。 | 背景を雨に濡れた夜のネオン街に変え、濡れたアスファルトの反射やぼやけたネオンサインを描きます。 |
| Switch to Image Edit = false | Switch to Image Edit = true | |
 |  | |
| 「HiDream-O1/」7102_hidream_o1_dev_simple.json | 画像サイズ：2048x2048 | |
| 7010 Text to Image(full) | 7010 SubGraph |
 |  |
| 「HiDream-O1/」7010_hidream_o1_full_t2i.json | |
| 英語プロンプト | 日本語プロンプト |
|
Magazine cover titled "HiDream-O1" vol.64. A central portrait of a young Japanese woman with long, wavy honey-brown hair and a neutral, elegant expression. She wears a two-piece ensemble made of pale pink chiffon: a cropped, off-the-shoulder top with voluminous sheer lantern sleeves and a matching high-waisted flared mini skirt. The fabric is translucent, catching soft light. She stands in a bright interior before a large white-framed window. Delicate cherry blossom branches with pale pink petals frame the sides of the composition. &br:Lighting is soft natural daylight, creating high-key tones and gentle shadows. Low depth of field with a blurred background. Text layout includes: "HiDream-O1 vol.64" in black serif typography at the top; "2026 春の絶対ルール" in red; "透け感とパステル" in blue; "可愛い Cropped mini dress" in pink and white italics; "肌も心もリセット" in teal; and "新年度、新しい自分へ" in bold white characters with a subtle drop shadow at the bottom. Sharp focus on the subject's face and garment texture. |
雑誌「HiDream-O1」第64号の表紙。中央には、ウェーブがかった長いハニーブラウンの髪と、ニュートラルで上品な表情をした若い日本人女性のポートレート。彼女はペールピンクのシフォン素材のセットアップを着用：ボリュームのあるシアーなランタンスリーブのオフショルダー・クロップドトップスと、お揃いのハイウエスト・フレアミニスカート。生地は半透明で、柔らかな光を捉えている。彼女は大きな白い窓枠の前の明るい室内に立っている。淡いピンクの花びらを持つ繊細な桜の枝が、構図の両端を縁取っている。 ライティングは柔らかな自然光で、ハイキーなトーンと穏やかな影を生み出している。背景をぼかした浅い被写界深度。テキストレイアウトには以下を含む：上部に黒のセリフ体で「HiDream-O1 vol.64」；赤色で「2026 春の絶対ルール」；青色で「透け感とパステル」；ピンクと白の斜体で「可愛い Cropped mini dress」；ティール（青緑）色で「肌も心もリセット」；そして下部に、かすかなドロップシャドウを付けた太い白文字の「新年度、新しい自分へ」。被写体の顔と衣服の質感にシャープな焦点を合わせている。 |
 |  |
| 画像サイズ：1728x2304 | |
| 7011 Image to Image(full) | 7011 SubGraph |
 |  |
| 「HiDream-O1/」7011_hidream_o1_full_i2i.json | |
| 英語プロンプト | 日本語プロンプト |
|
Close-up portrait of a Japanease young woman in her 20s, natural makeup, soft expression. 85mm lens, f/1.4, shallow depth of field, bokeh background. Soft window light from the left, golden hour glow, warm skin tones. Cinematic, realistic, high detail, professional photography. NOT: oversaturated, anime style, low resolution, distorted face. |
20代の日本の若い女性のクローズアップポートレート。ナチュラルメイク、柔らかな表情。 85mmレンズ、f/1.4、浅い被写界深度、背景のボケ。 左から差し込む柔らかな窓の光、ゴールデンアワーの輝き、温かみのある肌色。 映画のような、リアルな、細部までこだわった、プロフェッショナルな写真。 注意点：彩度過度、アニメ調、低解像度、歪んだ顔。 |
 |  |
| 画像サイズ：1728x2304 | |
| 7110 Text to Image(dev) | 7110 SubGraph |
 |  |
| 「HiDream-O1/」7110_hidream_o1_dev_t2i.json | |
| 英語プロンプト | 日本語プロンプト |
|
Magazine cover titled "HiDream-O1" vol.64. A central portrait of a young Japanese woman with long, wavy honey-brown hair and a neutral, elegant expression. She wears a two-piece ensemble made of pale pink chiffon: a cropped, off-the-shoulder top with voluminous sheer lantern sleeves and a matching high-waisted flared mini skirt. The fabric is translucent, catching soft light. She stands in a bright interior before a large white-framed window. Delicate cherry blossom branches with pale pink petals frame the sides of the composition. &br:Lighting is soft natural daylight, creating high-key tones and gentle shadows. Low depth of field with a blurred background. Text layout includes: "HiDream-O1 vol.64" in black serif typography at the top; "2026 春の絶対ルール" in red; "透け感とパステル" in blue; "可愛い Cropped mini dress" in pink and white italics; "肌も心もリセット" in teal; and "新年度、新しい自分へ" in bold white characters with a subtle drop shadow at the bottom. Sharp focus on the subject's face and garment texture. |
雑誌「HiDream-O1」第64号の表紙。中央には、ウェーブがかった長いハニーブラウンの髪と、ニュートラルで上品な表情をした若い日本人女性のポートレート。彼女はペールピンクのシフォン素材のセットアップを着用：ボリュームのあるシアーなランタンスリーブのオフショルダー・クロップドトップスと、お揃いのハイウエスト・フレアミニスカート。生地は半透明で、柔らかな光を捉えている。彼女は大きな白い窓枠の前の明るい室内に立っている。淡いピンクの花びらを持つ繊細な桜の枝が、構図の両端を縁取っている。 ライティングは柔らかな自然光で、ハイキーなトーンと穏やかな影を生み出している。背景をぼかした浅い被写界深度。テキストレイアウトには以下を含む：上部に黒のセリフ体で「HiDream-O1 vol.64」；赤色で「2026 春の絶対ルール」；青色で「透け感とパステル」；ピンクと白の斜体で「可愛い Cropped mini dress」；ティール（青緑）色で「肌も心もリセット」；そして下部に、かすかなドロップシャドウを付けた太い白文字の「新年度、新しい自分へ」。被写体の顔と衣服の質感にシャープな焦点を合わせている。 |
 |  |
| 画像サイズ：1728x2304 | |
| 7111 Image to Image(dev) | 7111 SubGraph |
 |  |
| 「HiDream-O1/」7111_hidream_o1_dev_i2i.json | |
| 英語プロンプト | 日本語プロンプト |
|
Magazine cover titled "HiDream-O1" vol.64. A central portrait of a young Japanese woman with long, wavy honey-brown hair and a neutral, elegant expression. She wears a two-piece ensemble made of pale pink chiffon: a cropped, off-the-shoulder top with voluminous sheer lantern sleeves and a matching high-waisted flared mini skirt. The fabric is translucent, catching soft light. She stands in a bright interior before a large white-framed window. Delicate cherry blossom branches with pale pink petals frame the sides of the composition. &br:Lighting is soft natural daylight, creating high-key tones and gentle shadows. Low depth of field with a blurred background. Text layout includes: "HiDream-O1 vol.64" in black serif typography at the top; "2026 春の絶対ルール" in red; "透け感とパステル" in blue; "可愛い Cropped mini dress" in pink and white italics; "肌も心もリセット" in teal; and "新年度、新しい自分へ" in bold white characters with a subtle drop shadow at the bottom. Sharp focus on the subject's face and garment texture. |
雑誌「HiDream-O1」第64号の表紙。中央には、ウェーブがかった長いハニーブラウンの髪と、ニュートラルで上品な表情をした若い日本人女性のポートレート。彼女はペールピンクのシフォン素材のセットアップを着用：ボリュームのあるシアーなランタンスリーブのオフショルダー・クロップドトップスと、お揃いのハイウエスト・フレアミニスカート。生地は半透明で、柔らかな光を捉えている。彼女は大きな白い窓枠の前の明るい室内に立っている。淡いピンクの花びらを持つ繊細な桜の枝が、構図の両端を縁取っている。 ライティングは柔らかな自然光で、ハイキーなトーンと穏やかな影を生み出している。背景をぼかした浅い被写界深度。テキストレイアウトには以下を含む：上部に黒のセリフ体で「HiDream-O1 vol.64」；赤色で「2026 春の絶対ルール」；青色で「透け感とパステル」；ピンクと白の斜体で「可愛い Cropped mini dress」；ティール（青緑）色で「肌も心もリセット」；そして下部に、かすかなドロップシャドウを付けた太い白文字の「新年度、新しい自分へ」。被写体の顔と衣服の質感にシャープな焦点を合わせている。 |
 |  |
| 画像サイズ：1728x2304 | |
| 英語 | 日本語 |
 |  |
| A beautiful Japanese woman in her 20s, natural facial features, smooth fair skin, long straight black hair, calm and confident expression, standing in a minimal modern interior. Medium shot, centered composition, soft diffused daylight from a side window creating gentle shadows, neutral color palette with subtle warm tones, sharp focus on eyes, shallow depth of field, clean background with slight blur, realistic skin texture, fine hair detail. | 20代の美しい日本人女性、自然な顔立ち、なめらかで明るい肌、長くまっすぐな黒髪、落ち着いて自信のある表情、ミニマルな現代的室内に立っている。ミディアムショット、中央構図、窓からの柔らかな拡散自然光で穏やかな影、ニュートラルでわずかに暖かみのある色調、目にシャープなピント、浅い被写界深度、背景はクリーンで軽くぼかし、リアルな肌質感と細かな髪のディテール。 |
| 英語 | 日本語 |
 |  |
| A beautiful Japanese woman in her 20s, natural facial features, smooth fair skin, long straight black hair, calm and confident expression. Bust-up portrait, centered composition, facing slightly toward camera, shoulders and upper chest visible. Soft diffused daylight from a side window creating gentle shadows and natural highlights on skin, neutral background with subtle blur, shallow depth of field, sharp focus on eyes, realistic skin texture, fine hair detail, balanced warm color tones. | 20代の美しい日本人女性、自然な顔立ち、なめらかで明るい肌、長くまっすぐな黒髪、落ち着いて自信のある表情。バストアップのポートレート、中央構図、ややカメラに向いた姿勢で肩から胸上までが見える。窓からの柔らかな拡散光で穏やかな影と自然なハイライト、背景はわずかにぼかしたニュートラルな空間、浅い被写界深度、目にシャープなピント、リアルな肌質感と繊細な髪のディテール、バランスの取れた暖色系の色調。 |
| 英語 | 日本語 |
 |  |
| A Japanese woman in her 20s drinking coffee at a café at night, natural facial features, smooth fair skin, long straight black hair, relaxed and thoughtful expression. Bust-up portrait, seated at a small table, holding a cup near her lips. Warm ambient lighting from overhead lamps and soft reflections from nearby windows, cinematic contrast between warm interior light and cool night tones outside. Shallow depth of field, blurred background with faint silhouettes of furniture and window lights, sharp focus on eyes and cup, realistic skin texture, subtle highlights and shadows. | 夜のカフェでコーヒーを飲む20代の日本人女性、自然な顔立ち、なめらかで明るい肌、長くまっすぐな黒髪、リラックスして思慮深い表情。バストアップのポートレート、小さなテーブルに座りカップを口元に持つ。天井照明の暖かい環境光と窓の反射による柔らかな光、室内の暖色と外の夜の冷たい色調の対比。浅い被写界深度、背景は家具や窓の光がぼんやりと見える、目とカップにシャープなピント、リアルな肌質感と繊細なハイライトと影。 |
| 英語 | 日本語 |
 |  |
| A tall, elegant Japanese businesswoman in her late 20s to early 30s, refined facial features, smooth fair skin, long straight black hair neatly styled, confident and composed expression. Medium-full portrait, standing posture emphasizing height, one hand lightly holding a tablet or folder. Wearing a tailored dark suit with a crisp white blouse. Clean modern office interior with glass and soft reflections. Soft directional lighting from large windows creating balanced highlights and gentle shadows, neutral cool color palette, shallow depth of field, sharp focus on face, subtle background blur, realistic fabric texture and fine details. | 背の高いエレガントな20代後半〜30代前半の日本人ビジネスウーマン、洗練された顔立ち、なめらかで明るい肌、整えられた長い黒髪、落ち着きと自信のある表情。身長が際立つミディアム〜全身寄りの構図、立ち姿で片手にタブレットや書類を軽く持つ。ダークカラーの仕立ての良いスーツに白いブラウス。ガラスや反射のあるモダンなオフィス空間。大きな窓からの柔らかな指向性光で均整の取れたハイライトと穏やかな影、ニュートラルでややクールな色調、浅い被写界深度、顔にシャープなピント、背景は軽くぼかし、リアルな生地感と細部描写。 |
| 英語 | 日本語 |
 |  |
| A mysterious 27-year-old professional female model, Japanese, refined facial features, smooth fair skin, long sleek black hair, subtle intense gaze with a composed expression. Bust-up portrait, slightly angled pose, one shoulder forward. Low-key lighting with soft directional light from the side creating deep shadows and dramatic contrast, dark minimal background with faint gradients. Shallow depth of field, sharp focus on eyes, soft shadow contours on face, realistic skin texture, controlled highlights, muted color palette with cool tones. | ミステリアスな27歳の日本人女性プロモデル、洗練された顔立ち、なめらかで明るい肌、長く艶のある黒髪、控えめで鋭い視線と落ち着いた表情。バストアップのポートレート、やや角度をつけたポーズで片肩を前に出す。サイドからの柔らかな指向性ライトによるローキー照明で深い影とドラマチックなコントラスト、背景は暗くミニマルでわずかなグラデーション。浅い被写界深度、目にシャープなピント、顔の陰影を柔らかく表現、リアルな肌質感、抑えたハイライトとクールトーン中心の色調。 |
| 英語 | 日本語 |
 |  |
| プロンプト → z-image の例 | |
| 英語 | 日本語 |
 |  |
| プロンプト → z-image の例 | |
| 英語 | 日本語 |
 |  |
| プロンプト → FLUX.2 の例 | |
| 英語 | 日本語 |
 |  |
| プロンプト → FLUX.2 の例 | |
| 英語 | 日本語 |
 |  |
 |  |
 |  |
| プロンプト → z-image の例 | |
| 英語 | 日本語 |
 |  |
| プロンプト → FLUX.1 の例 | |
| HiDream-O1 モデルは「ERNIE-Image」「FLUX2」などに比較すると日本語プロンプトへの追従が弱い用なので、英語のプロンプトで検証する |
| ① AURA Skincare Brand Visual System | ② Luxury Perfume Sand Ad Photography |
 2304x1728 |  1728x2304 |
|
This is a widescreen brand Visual Identity (VI) system design mockup presentation, arranged in a 3×3 grid layout, showcasing the complete and consistent visual system of a premium skincare brand named "AURA." The overall scene is set against a light gray seamless background, lit with soft diffused studio lighting and subtle, delicate shadows, conveying a clean, modern, minimalist, and high-end aesthetic. The brand's unified color palette consists of sage green, warm beige, and charcoal gray. There are no human figures in the image, nor any real-world brand names or watermarks. The specific contents of each grid are as follows: Grid 1 (top-left): A paper card mockup displaying the primary brand logo. A warm beige card with a subtle embossed texture is centered, with the main logo text 'AURA' printed in charcoal gray on the front, along with the smaller subtitle 'BOTANICAL SKINCARE' below. Grid 2 (top-center): Secondary marks and monogram variants. The background is sage green, with a minimalist geometric uppercase letter 'A' mark displayed in the center, accompanied by an array of patterns derived from the letter arranged alongside it. Grid 3 (top-right): Color swatches and typography poster. Three square color swatches (sage green, warm beige, charcoal gray) are neatly arranged at the top of the poster, with typography samples displayed below, including a black headline 'Typography,' a subtitle 'Helvetica Now Display,' and a clear sample text reading 'The essence of pure, natural beauty.' Grid 4 (middle-left): A packaging box mockup for a serum product. A sage green paper box stands upright, with the white 'AURA' logo printed on the front, 'Glow Serum' written below, and the volume information '30ml / 1 fl oz' printed at the bottom. Grid 5 (center): A hero visual photograph of a moisturizer product. A warm beige matte ceramic cream jar is placed on a minimalist geometric pedestal, with 'AURA' and 'Daily Moisture' printed in charcoal gray on the jar body. Grid 6 (middle-right): A set of three body lotion pump bottles in different sizes. Three sage green frosted-finish pump bottles stand side by side in ascending height order, uniformly printed with 'AURA' and 'Body Loti |
A high-end luxury product advertising photograph, utilizing a horizontal composition and shallow depth of field. From a direct top-down overhead perspective, a premium perfume bottle lies flat on a cream-colored fine sand surface with delicate texture. The perfume bottle features a thick, transparent rectangular glass body containing a pale golden liquid, topped with a smooth gold metallic cylindrical cap. The front of the perfume bottle is fully visible and positioned upright, with a minimalist white square label affixed to the center. The label bears clearly printed black serif text reading 'L'Éternité', beneath which are two smaller lines of black sans-serif text reading 'EAU DE PARFUM' and '100 ml - 3.4 FL.OZ.' respectively. The cream-colored fine sand surface surrounding the perfume displays gentle, undulating wave-like traces suggesting recent breeze or manual interaction. The perfume is placed at the visual center of the frame, with generous blank space intentionally left around it. In the blank area of the lower right corner, a small dry branch bearing two half-open white magnolia blossoms and several emerald-green leaves is elegantly placed; in the upper left corner, three pristine white magnolia petals are sparsely scattered as supplementary visual elements, naturally enhancing the richness of the scene. A warm, bright beam of natural sunlight streams diagonally from the upper left corner of the frame, casting clear and realistic warm-toned shadows on the lower right side of the sand undulations, the perfume bottle, and the magnolia flowers. Under the sunlight, the transparent glass bottle body creates bright golden caustic light patterns on the sand surface. The overall image presents an extremely high degree of photorealism at a 3D rendering level, with the focal point sharply concentrated on the front of the perfume bottle, while the scattered petals at the edges and sand surface textures exhibit a delicate bokeh transition. The overall lighting is sophisticated, warm, and serene. |
| ③ Logo Miniature Hyper-Realistic Photography | ④ Night Street Looking-Back Portrait Photography |
 1728x2304 |  2304x1728 |
| A vertical ultra-realistic miniature photography work. At the center of the frame is a massive 3D physical logo with the English text 'BRAND NAME' clearly and completely engraved on its surface. The logo retains its original shape and vivid colors. Around and on the surface of the logo, there are numerous realistic miniature human figurines at approximately 1:50 scale engaging in various interactions. Some figurines stand on scaffolding, painting the edges of the logo with brushes; some climb the logo structure using ladders and ropes; others wipe the surface clean with cloths, while some hold miniature cameras to take photos. The figurines are equipped with props such as ladders, ropes, and buckets. The background is a clean, pure white, non-gradient studio backdrop. The shot is captured from a macro photography perspective with shallow depth of field and a slight overhead angle to emphasize the massive sense of volume. The lighting consists of soft diffused white light from above and fill light from the side, casting realistic and delicate tiny shadows on the pure white ground and around the logo. The colors are true to life, and the text 'BRAND NAME' is crisp and sharp. |
A realistic-style nighttime street portrait photography shot. The image employs a horizontal composition, with the main subject being a young woman standing in a city alleyway at night, her body slightly turned away from the camera, glancing back over her shoulder to gaze into the lens. Her facial expression is soft and natural, with slightly tousled dark long hair cascading freely, a few strands falling and resting against both sides of her cheeks, framing her facial contours. She wears a thick, textured dark coat. Due to the shallow depth-of-field effect achieved with a wide aperture, the city alleyway in the background is deeply blurred. The background is filled with colorful circular bokeh, formed by distant neon lights and streetlamp halos. The lighting in the image presents a cinematic quality, with warm red and yellow light blending together, illuminating the woman's profile, the bridge of her nose, and the edges of her hair, creating soft highlights and distinct rim lighting; meanwhile, the ambient light in the surrounding background carries cool tones, with warm and cool colors intermingling throughout the frame, creating a dim yet emotionally evocative nighttime atmosphere. The photograph possesses extremely high clarity and hyper-realistic quality—the woman's skin texture, subtle facial contours, reflections in her eyes, and the sheen of every individual strand of hair are all rendered with vivid realism, the overall visual effect resembling a cinematic still frame captured through a 35mm lens. |
| ⑤ Narrow House Isometric Cross-Section Render | ⑥ Ghibli Style |
 1728x2304 |  1728x2304 |
|
A 3D architectural interior rendering with a cross-section effect, using an isometric view to overlook the complete floor plan of a narrow elongated residential unit, presenting an extremely high level of photorealistic quality. The image is in vertical composition, with no roof, and the walls are cut flat at the top, exposing the white wall thickness, to clearly display the internal spatial layout. Bright, soft natural light streams in diagonally from windows on the right side, casting realistic light and shadow distributions on the floors and furniture, with refined and lifelike global illumination and material textures. The interior space is divided from the bottom to the top of the image into three main living zones: At the bottom of the image is the entrance foyer and living room area, with floors laid in warm-toned natural wood herringbone parquet. In the lower right corner is a dark gray security entrance door, with a minimalist shoe-changing bench and a coat rack placed against the adjacent wall. At the center of the living room is an L-shaped light gray fabric sofa, with a woven rug featuring black-and-white geometric patterns beneath it, and a double-tiered round natural wood coffee table in front, casually topped with a few magazines and a cup of coffee. On the left wall, a black flat-screen TV is wall-mounted, with a white floating TV console below it. The right wall features a large floor-to-ceiling glass window, with a lush, large Monstera deliciosa plant placed in the corner by the window. The middle of the image transitions smoothly into an open-plan kitchen and dining area, where the floor material changes from wood flooring to light gray matte anti-slip ceramic tiles. Along the left wall is an L-shaped modern pure white acrylic cabinet system with a quartz stone countertop, with a built-in stainless steel oven and microwave. On the countertop are a single-basin sink, an Italian espresso machine, several knives, and black spice jars, with a silver double-door refrigerator standing at the far left. Slightly to the right is a rectangular walnut dining table, paired with four modern-style black leather dining chairs, with a transparent glass vase filled with white lisianthus (Eustoma) placed at the center of the dining table. At the top of the image is the relatively private resting and bathroom area. The upper left contains a spacious bedroom, with the floor returning to wood flooring. At the center of the room is a comfortable double bed, dressed with pure white bed sheets, fluffy pillows, and a navy blue knitted bed-end throw blanket. On each side of the bed is a wooden nightstand with drawers, each topped with a warm yellow metallic desk lamp. On the left wall of the bedroom, a full row of floor-to-ceiling louvered-door wardrobes is custom-fitted. The upper right space is a modern bathroom with wet-dry separation, with floors laid in black-and-white hexagonal mosaic tiles. Inside the bathroom are a shower enclosure with transparent glass partition (with a silver showerhead), a white wall-mounted smart toilet, and a floating oak wash | A portrait-oriented animation illustration rendered in the classic Studio Ghibli art style. At the center of the scene, an ancient stone stairway winds upward, its surface weathered and covered with thick emerald-green moss and a few small white wildflowers. A young girl with short black hair stands midway up the stone steps, wearing a light blue vintage dress with a white Peter Pan collar and carrying a worn brown leather satchel on a cross-body strap. She tilts her head slightly upward, her expression conveying curiosity and wonder as she gazes at the forest above. In the foreground grass to the girl's left, a mysterious fantastical small creature hides—round and plump, entirely snow-white, with two long ears—holding up a giant green taro leaf as an umbrella. The background is a lush, vibrant ancient forest. Several enormous old-growth trees, so wide they would require several people to encircle, rise from the ground with clearly defined bark textures, their surfaces draped with thick vines and fern plants. The background foliage is rendered in distinct layers, exhibiting the translucent quality of traditional watercolor painting with an extraordinarily rich range of green tones. In the upper portion of the image, bright sunlight penetrates the dense canopy, forming several distinct golden Tyndall beams that gently fall upon the girl's skirt hem and the moss-covered stone steps, creating vivid dappled lighting effects. Through gaps in the leaves at the very top of the frame, a deep blue sky is faintly visible at a great height, along with a single massive, cotton-candy-like midsummer cumulus cloud with a strong sense of volume. The overall art style conveys a rich sense of summer healing and a fantastical fairy-tale atmosphere. The characters and creatures are rendered using delicate cel-shading flat coloring techniques with soft edge lines, seamlessly blending with the hand-painted watercolor-textured, high-detail background. |
| ⑦ Fashion Figure Outfit Structure Breakdown | |
 3104x1312 | |
|
A hand-drawn style 2D illustration presenting a fashion concept breakdown sheet. The composition is in a wide-format layout with a beige parchment-textured background, clean and soft lighting, and a strong design sketch atmosphere. At the center of the image is a fashionable and confident young woman shown in full-body view. She has slightly wavy long hair and wears a camel-colored matte leather trench coat, layered over a black lace shapewear bodysuit and sheer tulle leggings. Her pose is natural yet dynamic, exuding a subtly sensual quality. Radiating outward around her are structured elements and bilingual Chinese-English handwritten-style labels: Upper left is the garment layer breakdown: displaying the lingerie components, with 'Shapewear & Lingerie / シェイプウェア＆ランジェリー' written alongside, leading to a material annotation 'soft lace / ソフトレース'. The trench coat is labeled 'Trench Coat / トレンチコート' with 'matte leather / マットレザー'. The leggings are labeled 'Leggings / レギンス' with 'tulle material / チュール素材'. Upper right is the expression board: containing four expression headshots, labeled respectively 'Neutral / ニュートラル', 'Shy / 恥じらい', 'Surprised / 驚き', and 'Focused / 集中'. Right middle section features close-up shots: a magnified display of fabric texture labeled 'Fabric Folds / 生地の折り目', hand poses labeled 'Hand Gesture / 手のジェスチャー', and a partial skin detail labeled 'Skin Texture / 肌の質感'. The bottom section shows lifestyle accessories: an open handbag labeled 'Handbag / ハンドバッグ' with everyday items scattered around. A lipstick labeled 'Lipstick / 口红' with 'shade #520 / シェード#52', perfume labeled 'Perfume / 香水', a compact labeled 'Compact Powder / コンパクトパウダー', hand cream labeled 'Hand Cream / ハンドクリーム', a diary labeled 'Diary / 日記', and a health supplement bottle labeled 'Supplements / サプリメント'. | |
| ⑧ First-Person City Night Street Photography | |
 1312x3104 | A first-person perspective real-world urban nightscape photography shot, composed in vertical/portrait orientation to emphasize the depth of the street and the towering height of the buildings on both sides. The image depicts a narrow, bustling, and rain-slicked modern city street, with the line of sight extending from the center of the street toward the distant vanishing point. Soaring buildings on both sides block most of the night sky; their exterior walls are covered with intricate metal fire escapes, rusted pipes, and densely packed air conditioning units, with some windows emitting warm yellow light. The street is flanked by a dense array of glowing neon signs: on the left foreground, a protruding red vertical sign clearly reads 'ホテル' in glowing English letters; on the right building façade hangs a bright blue rectangular sign with white text reading 'BAR', and directly beneath it is a small yellow square lightbox with black text reading '24時間営業'. Further down the street, atop a high-rise, a massive electronic billboard displays dazzling cyan glowing text reading 'ネオン シティ'. The ground is rough black asphalt; the rain-soaked surface has multiple uneven puddles that act like mirrors, perfectly reflecting the red and blue glows of the signs above. In the dead center of the road, a vintage classic yellow taxi is driving deeper into the street, its red tail lights glowing, with the yellow roof-mounted sign clearly displaying the black text 'TAXI'. On the right sidewalk edge, two pedestrians wearing dark trench coats and holding black umbrellas walk with their backs to the camera, heading further down the street. At the vanishing point, in the narrow "canyon" gap formed between the tall buildings on both sides, a massive needle-shaped steel television tower looms faintly through the night sky and thin mist, its spire blinking with a faint red aviation obstruction light. The lighting in the image has a strong cinematic volumetric light effect, with water mist permeating the air. The overall color palette presents a stark contrast between cool cyan-blue tones and intense orange-red tones, making the image highly immersive and photorealistic. |
| HiDream-O1 モデルは「ERNIE-Image」「FLUX2」などに比較すると日本語プロンプトへの追従が弱い用なので、英語のプロンプトで検証する |
| 生成画像 | プロンプト |
 1728x2304 | Split-screen conceptual poster with vertical split composition, left half representing the present and future featuring a solid stone analog clock dial with bright red hands and red scale markings, set against a deeply cracked, dry, parched earth texture in muted cool gray tones, embodying slow, static structural decay and temporal erosion, right half representing the past that is violently disintegrating, dissolving and exploding into chaotic dust, debris, flying stone shards and swirling cosmic nebula energy with a deep blood red and black color palette, the circular clock frame itself is half solid weathered stone and half crumbling into particles merging the two worlds, with the red clock hands fully positioned on the left present/future side of the dial, embodying the concept of time, past vs future, slow decay vs violent collapse, hyper-realistic 3D render with cinematic dramatic lighting, high contrast, sharp details on the left side, motion blur and dynamic particle effects on the right side, photorealistic textures of cracked earth and floating dust, atmospheric haze, futuristic graphic design with minimalist red vertical UI elements and custom technical text overlays in the corners, bold red stylized title text "ETERNAL DAWN" in the top right corner, small red technical metadata text "SPECDARY BOTHUNG" directly below the main title in the top right, dense lines of small red technical data blocks including "TIMESTAMP: 00:00:00", "COLLAPSE RATE: 99.8%", "TEMPORAL ANOMALY DETECTED", "PAST DIMENSION: DECAYING" in the bottom right corner with a red horizontal accent bar and footer text "FOR DECAY OF THE EONS" below, small red header text "TEMPORAL SHIFT PROTOCOL" with vertical red accent line running down the left edge in the top left, vertical red accent line with small red technical text annotations including "PRESENT: STABLE", "FUTURE: UNFOLDING" along it in the bottom left, red text "WARNING: TEMPORAL DISINTEGRATION" curving around the center clock dial perimeter, small red footer text "PAST FADES, FUTURE RISES" spanning the split at the bottom center, shot in the style of a high-end sci-fi movie poster, 8K ultra-high resolution, photorealistic, cinematic composition, dramatic depth, moody and intense atmosphere, detailed particle simulation, photorealistic material rendering. |
 1728x2304 | Please create a poster image themed around a traditional Japanese summer festival. Place the words "大江戸 夏祭り" in a bold, calligraphic Japanese font in the center of the poster, with the words "2026年 開催" in a Mincho font below it. Beautifully depict lanterns and fireworks in the background. Ensure the text is legible and avoids meaningless symbols. |
 2304x1728 | Close-up portrait of a smiling young East Asian woman with long dark hair and bangs, wearing a pink camisole and a gold pendant necklace. She is seated at an outdoor cafe at night. In the blurred background, a vibrant blue neon sign displays the Japanese text "美女茶屋って出来るかな?cafe". The lighting is cinematic with warm bokeh from street lights and a cool blue glow from the neon sign. The composition is centered with a shallow depth of field, highlighting the subject's clear skin and expressive eyes. |
 2304x1728 | A photorealistic 1:1 aspect ratio image, a close-up of a freshly made, steaming omelet rice placed on a table in a traditional coffee shop. The words "殻を破ろう" are handwritten in bright red ketchup on top of the vibrant yellow egg. The image realistically reproduces the three-dimensional thickness of the ketchup and the almost dripping texture at the end of the letters. |
 2304x1728 |
A mockup of a high-quality matcha packaging design. The box should have a calm, deep green color and a texture reminiscent of Japanese washi paper. The words "極上 宇治抹茶" should be written vertically in the center, with "京都の伝統" below it in smaller, clearly legible Japanese text. |
 2304x1728 | Please generate an image of a handwritten-style chalkboard pop-up ad with the Japanese text, "新鮮なオーガニック りんご - 本日半額！", written precisely in a chalk art style. Place a blurred pile of apples in the background. |
 2304x1728 | A photorealistic image with a 1:1 aspect ratio, depicting a view of the snowy landscape outside through a condensation-covered glass window from inside an old Japanese house in winter. The words "春は すぐそこ" are traced with a finger on the cloudy white glass, and only the traced letters are transparent, allowing a clear view of the red camellia flowers blooming outside. |
 2048x2048 | A photorealistic 1:1 aspect ratio image of moss-covered, old soil in the grounds of a Japanese temple after a rain shower, viewed from directly above. Large characters reading "雨降って 地固まる" are visible on the surface of the rain-soaked, dark soil, written with a thick brush dipped in water. The water-written characters reflect light even more strongly than the surrounding wet soil, depicting a subtle blurring effect. |
 1728x2304 | A vertically oriented photographic triptych with a 35mm film photography aesthetic. The frame is divided vertically into three equally sized panels stacked top to bottom, each depicting different poses of the same young woman outdoors in sunlight. The woman has dark, slightly tousled wavy hair with a few strands naturally lifted by a gentle breeze, and she wears a simple white camisole top. The background across all three panels is a clear, bright, cloudless azure sky. The upper panel employs a low-angle upward shot; the woman looks down quietly at the camera with a direct yet slightly contemplative gaze, bright afternoon sunlight illuminating one side of her face and shoulder from the side, creating soft, natural shadows. The middle panel uses an eye-level perspective with a slight upward tilt; the subject is positioned slightly off-center to emphasize the expansive sky, her head turned toward the right side of the frame, gazing into the distance with a pensive and detached expression; a small brownish columnar rooftop structure is faintly visible at the lower-left edge of the frame. The lower panel is an extreme close-up side-profile shot of her face; the woman has her eyes gently closed, her head tilted slightly back, her expression supremely calm and serene, the bright sunlight precisely delineating the elegant contours of her facial features and the line of her neck. The overall imagery is shot with a wide aperture, with sharp focus on the subject's face and a shallow depth of field. In terms of color palette, warm and authentic natural skin tones contrast with the vivid, bright blue sky. The images exhibit pronounced film grain, a nostalgic atmosphere, and the style of high-end editorial magazine portraiture. The lighting is purely natural diffused sunlight with no artificial lighting or digital skin-retouching, preserving the skin's genuine texture and minor imperfections. The overall mood is tranquil, ethereal, fresh, and imbued with a remarkable sense of airiness. |
 2304x1728 | First-person vertical real-world urban night photo. Narrow, wet, busy city street with strong depth; towering buildings on both sides with fire escapes, pipes, AC units, warm window lights. Red vertical "HOTEL" neon on left, bright blue "BAR" sign and yellow "OPEN 24 HOURS" box on right. Distant huge digital billboard glowing cyan "NEON CITY". Rough wet asphalt reflects neon lights; vintage yellow taxi with "TAXI" sign drives down center, red taillights on. Two pedestrians in dark coats with black umbrellas walk away on right sidewalk. Giant needle-like steel TV tower peeks through building gaps at end, red aircraft warning light glowing. Cinematic volumetric light, misty air, cool cyan-blue vs warm orange-red contrast, immersive and realistic. |
 2048x2048 | A charming, whimsical 3D illustration featuring a fluffy, plush-textured white duck sitting comfortably in a fuzzy coral-pink armchair, holding a bright red mug of steaming black coffee against a solid warm coral-red background; defined by its tactile, huggable fabric-like texture across all elements, a bold, minimalist warm color palette of creamy white, vibrant orange, and rich reds, gentle anthropomorphism that blends cuteness with relatable cozy relaxation, soft diffused lighting, and a subtle film grain that adds a nostalgic, handcrafted feel, creating a playful yet serene mood perfect for themes of calm downtime and morning routines. |
 2048x2048 |
Create a scene that includes multiple famous landmarks from around the world. The image must include: – Kinkaku-ji (the Golden Pavilion in Kyoto), with its gold-covered exterior reflected in a pond – The Leaning Tower of Pisa, clearly tilted at its characteristic angle – The Eiffel Tower in the background Arrange them in a single coherent scene with natural perspective and scale. Each landmark should be recognizable and accurately represented. Use a realistic or semi-realistic style. Avoid incorrect shapes or distorted structures |
 1728x2304 | A high-precision, cinematic-quality close-up portrait photograph. At the center of the frame is a woman with long, golden curly hair and refined features. She is in the midst of a genuine, joyful burst of laughter, her head tilted slightly backward, eyes tightlA vertical Japanese film-style portrait photograph. A young East Asian woman turns to look back over her shoulder, her expression soft with a hint of melancholy, lips slightly parted. She has fair skin with natural, light makeup, wearing a dark jacket, with long, slightly tousled dark brown curly hair. The setting is a cluttered interior space (such as an old bookstore or studio), with a blurred background revealing wooden shelving, papers, and boxes. The foreground is partially obscured by translucent glass or plastic, creating a sense of depth and layering. Warm golden hour sunlight enters from the left, producing strong backlighting and a hair rim light, accompanied by lens flare, Tyndall light beams, and film grain. Shallow depth of field with precise focus on the face, the overall image evoking a cinematic, ethereal, and nostalgic atmosphere. |
 2304x1728 | Full-body editorial fashion shot, elegant Japanease woman aged 28-30 with soft natural blonde hair styled in a loose low bun, wearing a tailored soft cream ribbed knit long-sleeve maxi dress with a fitted waist and flowing floor-length skirt, minimalist nude leather mules and delicate gold jewelry, standing leaning against an ornate carved stone balustrade on a gravel terrace in the garden of a grand Lake Como villa, tall iconic Italian cypress trees framing the background, calm still Lake Como water stretching out to hazy misty Alpine mountains in the distance, overcast soft diffused natural light, muted warm earthy color palette, quiet luxury aesthetic, hyper-realistic editorial fashion photography, natural skin texture with subtle pores, soft bokeh foreground foliage, shallow depth of field, shot on a medium format camera with an 85mm f/1.4 lens, sharp focus on the subject, fine film grain, muted contrast, timeless sophisticated mood, shot by Lachlan Bailey in the style of Vogue editorial, 8K ultra-high resolution, photorealistic, cinematic composition following the rule of thirds, golden hour overcast lighting, atmospheric haze, refined Italian villa garden details. |
 1728x2304 |
A realistic-style nighttime street portrait photography shot. The image employs a horizontal composition, with the main subject being a young woman standing in a city alleyway at night, her body slightly turned away from the camera, glancing back over her shoulder to gaze into the lens. Her facial expression is soft and natural, with slightly tousled dark long hair cascading freely, a few strands falling and resting against both sides of her cheeks, framing her facial contours. She wears a thick, textured dark coat. Due to the shallow depth-of-field effect achieved with a wide aperture, the city alleyway in the background is deeply blurred. The background is filled with colorful circular bokeh, formed by distant neon lights and streetlamp halos. The lighting in the image presents a cinematic quality, with warm red and yellow light blending together, illuminating the woman's profile, the bridge of her nose, and the edges of her hair, creating soft highlights and distinct rim lighting; meanwhile, the ambient light in the surrounding background carries cool tones, with warm and cool colors intermingling throughout the frame, creating a dim yet emotionally evocative nighttime atmosphere. The photograph possesses extremely high clarity and hyper-realistic quality—the woman's skin texture, subtle facial contours, reflections in her eyes, and the sheen of every individual strand of hair are all rendered with vivid realism, the overall visual effect resembling a cinematic still frame captured through a 35mm lens. |
 1728x2304 | A hyper-realistic, cinematic-grade portrait photography work in vertical composition. At the center of the frame is a young japanease woman, presented in a three-quarter profile view with her face tilted slightly upward. She has black hair, the strands soft with a casually tousled feel. The woman wears a confident smile with a natural gaze, her arms crossed over her chest, dressed in a solid cyan long-sleeve crew-neck shirt. The background is a deep, pure black studio setting. The image exhibits dramatic lighting effects: from the right side, a bright red rim light is cast, precisely outlining the edges of her hair, facial contour, shoulders, and arms on the right; from the left side, a soft cool blue fill light illuminates her left cheek and clothing details. The red and blue dual-tone lighting creates a striking warm-cool contrast on the subject, greatly enhancing the three-dimensionality and cinematic atmosphere of the image. The rendering of the subject's face is extremely detailed, with skin texture, pores, and catchlights in the eyes all clearly and realistically depicted. |
 2304x1728 | A vertically formatted minimalist pictogram poster themed as a pour-over coffee guide. The background is a uniform clean off-white. The design employs a flat vector style, composed of highly simplified pictographic icons rendered in dark brown thick outlines with warm orange accent color blocks, free of gradients or 3D shadow effects, with a visual language as intuitive and legible as public signage. Centered at the top of the image is a line of dark brown bold sans-serif large title text: ‘ハンドドリップコーヒーガイド’. Below the title, four step blocks are evenly distributed along the vertical central axis, each block containing a large symbolic icon and accompanying text. The first block: on the left is a minimalist hand-crank coffee grinder graphic; on the right are two lines of text—the first line in bold orange reading ‘STEP 1’, the second in dark brown reading ‘コーヒー豆20gを挽く’. The second block: on the left is a slender-necked pour-over kettle graphic with a circular thermometer dial; the right-side text reads orange ‘STEP 2’ and dark brown ‘お湯を92℃まで沸かす’. The third block: on the left is a dynamic graphic depicting the kettle spout pouring water into a V-shaped filter dripper, with three wavy lines representing steam above the dripper; the right-side text reads orange ‘STEP 3′ and dark brown ’40gのお湯を注ぎ、30秒待つ’. The fourth block: on the left is a rounded coffee mug graphic with steam lines rising from the rim; the right-side text reads orange ‘STEP 4’ and dark brown ‘残りのお湯を注いで楽しむ’. At the very bottom of the image is a horizontal dark brown thin dashed line serving as a visual closure, with a smaller line of decorative text centered beneath it reading: ‘ENJOY YOUR DAILY COFFEE’. The overall composition is rigorously aligned with clear information hierarchy, perfectly suited to a vertical aspect ratio. |
 2304x1728 |
Educational comic book style infographic showing the coffee making process. The background has a light beige vintage paper texture. At the top center of the image is a bold brown title that reads 'コーヒーの製造工程', with a smaller English subtitle 'コーヒーの作り方' below. The main part consists of six step blocks connected by brown dotted arrows, arranged in two rows up and down, forming a Z-shaped visual guide. The first step in the upper left corner illustrates a coffee tree full of ripe red coffee cherries, with a cut coffee cherry next to it showing the beans inside, labeled 'ステップ1：チェリーの収穫' below the block. The second step in the upper middle shows wooden fermentation boxes filled with coffee beans, labeled 'ステップ2：パルプ除去と発酵'. The third step in the upper right depicts coffee beans being dried under the sun on bamboo mats, labeled 'ステップ3：天日干し'. The fourth step in the lower left shows a vintage metal roasting machine with coffee beans rolling inside and steam rising, labeled 'ステップ4：焙煎'. The fifth step in the lower middle features a stone grinder pouring out smooth coffee grounds, labeled 'ステップ5：粉砕'. The sixth step in the lower right shows a production line where coffee liquid is being poured into molds, with finished packaged coffee beside it, labeled 'ステップ6：抽出と成形'. The four corners of the image are decorated with hand-drawn coffee leaves and coffee beans. The overall color palette consists of warm brown, caramel, cream, deep red and olive green, with delicate lines and clear layout. |
 1728x2304 | A 6-panel comic page, 2 columns × 3 rows, black border between each panel.Panel 1 (top-left): wide shot — a young woman in a red coat stands at a rainy train station. Caption box: "彼女はこの瞬間を3年間待っていた。"Panel 2 (top-right): close-up on her face — anxious eyes, rain on her cheek. No text.Panel 3 (mid-left): medium shot — a train arrives, doors slide open, steam rising. Sound effect text: "ヒューッ"Panel 4 (mid-right): her point-of-view — a man in a gray jacket steps out, back to camera.Panel 5 (bottom-left): extreme close-up — her hand trembling as she reaches forward.Panel 6 (bottom-right): wide shot — they face each other under one umbrella. Caption box: "到着がすべてを変えることもある。"Ink and watercolor style, cool blue-gray palette, expressive line art. Same character design consistent across all 6 panels. |
 2304x1728 | Educational comic book infographic. Five vertical panels side by side, earthy color palette, hand-drawn illustration style. Top title: "北米の在来種". Style: comic line art, watercolor fills, white panel backgrounds, bold headers.Panel 1: Gray squirrel holding an acorn. Header: "東部ハイイロリス" Fact: "ドングリを埋めて森林の成長を助ける。" Arrow → tail: "ふさふさした尾"Panel 2: Robin with orange-red breast on a branch. Header: "アメリカコマドリ" Fact: "春の到来の兆し。" Arrow → breast: "オレンジがかった赤い胸"Panel 3: Red maple tree with autumn leaves. Header: "赤いカエデの木" Fact: "一年中野生動物の避難場所となる。" Arrow → leaves: "秋には赤くなる"Panel 4: Deer with white tail raised. Header: "オジロジカ" Fact: "白い尾は危険のサイン。." Arrow → tail: "警報信号"Panel 5: Monarch butterfly on a flower. Header: "オオカバマダラ" Fact: "毎年3,000マイル移動する。" Arrow → wings: "警告色" |
 2304x1728 | A sticker sheet display in digital illustration format. The composition is in landscape orientation with a light yellow polka-dot pattern background. The stickers are arranged in a 2-row by 3-column grid layout, showcasing 6 emoji stickers featuring the same 2D anime character. Each sticker has a thick white outline along its edges, giving a three-dimensional die-cut sticker effect. The consistent sticker character is a cute girl in a 2D anime style with long pink twin-tail hair, blunt bangs, white cat-ear hair accessories on top of her head, a classic blue sailor uniform, a large red bow tied at the chest, and deep emerald-green large eyes. In the upper-left sticker, the girl wears a bright smile with her right hand raised high in a greeting gesture, and the lower-left corner features the English word ‘HELLO’ in pink with a thick white outline. In the upper-middle sticker, the girl playfully winks with one eye while forming a heart shape with both hands in front of her chest, with three small floating red hearts around her and the red English word ‘LOVE’ at the bottom. In the upper-right sticker, the girl has large teardrops at the corners of her eyes, holding white tissues in both hands to wipe her tears with a pitiful, aggrieved expression, and the upper-right corner features the blue text ‘TAT’. In the lower-left sticker, the girl puffs out her cheeks with arms crossed over her chest in an angry pose, with a red anger symbol (💢) drawn to the right of her head, and the purple English word ‘HUMPH’ below. In the lower-middle sticker, the girl’s eyes are wide open, both hands covering her cheeks, mouth exaggeratedly open in an “O” shape showing an extremely shocked expression, with the orange English word ‘OMG’ in the upper-left corner. In the lower-right sticker, the girl laughs with her eyes closed, giving a thumbs-up with her right hand pointing forward, with yellow four-pointed sparkling star effects in the background, and the bright green English word ‘OK!’ at the bottom. The overall image has vibrant colors, primarily using soft macaron tones, with lighting and shading rendered in a Japanese cel-shading flat-color style, featuring clean and smooth character outlines full of vitality and fun. |
 1728x2304 |
Vertical rectangular language learning flashcard in traditional Japanese Ukiyo-e woodblock style. Whole card framed by dark blue traditional Seigaiha wave borders, with aged beige rough washi paper texture background. Upper two-thirds: Ukiyo-e illustration. A graceful spotted sika deer stands calmly in the center, with delicate black outlines and traditional flat coloring. Background: classic Japanese scenery — snow-capped Mount Fuji in distance, stylized layered clouds, falling pink cherry blossom petals. A red square antique seal with the Chinese character '鹿' in seal script at the lower right. Color scheme: indigo, vermilion, moss green, beige, with woodblock color misalignment and vintage charm. Lower one-third: centered text area. Top: small hiragana 'しか' large Japanese kanji "鹿" below it romaji 'shika' then English 'Deer'. Bottom: bilingual example — "日本語：鹿が桜の花を見ている" "英語：The deer is looking at the cherry blossoms." All black text with slight brush calligraphy texture, matching Ukiyo-e antique style. |
| model | hidream_o1_image_dev_fp8_scaled.safetensors, gemma4_e4b_it_fp8_scaled.safetensors |
| steps | 28 |
| cfg | 1.0 |
| seed | 270186383729385 |
| size (pixel) | 2048x2048, 1728x2304, 2304x1728, 3104x1312, 1312x3104 |
| Workflow | 7110_hidream_o1_dev_t2i.json |
| 画像ピクセル、テキストタグ、およびタスク条件を1つの共有連続空間に統合する、最先端の画像生成基盤モデル（例：HiDream-O1-Imageなど）のアーキテクチャ |
| AIモデルにおける「fp8」と「fp8_scaled」の主な違いは、量子化（データの圧縮）を行う際に「スケール係数」を用いて精度低下を防ぐ工夫がされているかどうか |
| 特徴 | FP8 | FP8_scaled |
| VRAM節約 | 大幅に節約可能 | 大幅に節約可能 |
| 生成品質・精度 | 誤差がやや出やすい | 品質劣化が抑えられ、より元のモデルに近い結果が出やすい |
| 処理速度 | 高速 | 高速（対応GPUでは最適化によりFP8より速い場合もある） |
| AIモデルにおける「FP8」と「MXFP8」の最大の違いは、数値データの拡大縮小（スケーリング）を行う範囲の細かさ。MXFP8はFP8の進化版であり、メモリ使用量を抑えつつ、モデルの推論・学習精度をより高く維持できる |
| 項目 | FP8 | MXFP8 |
| スケーリング（拡大縮小）の粒度 | テンソル（データのかたまり）全体に対して「1つの共通のスケール値」を適用する。そのため、非常に大きな値と小さな値が混在するデータでは、一部の精度が落ちることがある | データを 32個などの「小さなブロック」に分割し、ブロックごとに異なるスケール値を適用（マイクロスケーリング）する |
| 表現できる精度と範囲 | データ全体でスケールを合わせる必要があるため、精度を優先する形式（E4M3）と、表現できる数値の範囲を優先する形式（E5M2）を使い分ける必要がある | ブロックごとに細かくスケールを調整できるため、精度が高いE4M3形式をネットワーク全体で使いやすくなり、数値の表現力と安定性が向上しする |
| ハードウェアの対応 | NVIDIA の Hopperアーキテクチャ（H100など）や Ada Lovelace 世代などでサポートされている | NVIDIA の Blackwellアーキテクチャ以降で本格的にサポートされる最新のフォーマット |
