更快速的視覺迭代
使用 Flare 快速探索創作方向並比較不同變化。只需根據場景需求選擇合適的畫質與複雜度，而不必在每次請求中反覆堆疊細節。
GPT Image 2.6
創作更快速。精雕每個細節。讓你的圖像栩栩如生。
GPT Image 2.6
了解 GPT Image 2.6
GPT Image 2.6 為圖像創作與編輯帶來更快速的迭代體驗。Flare 專為快速視覺探索而設計，而 Sunburst 則專注於細膩的構圖與跨編輯的高度一致性。無論是構建場景、微調產品，或是將角色帶入全新背景，都能更精準地掌控關鍵細節。
此模型系列也在 Image API 中引入了動態 GIF 與 SVG 輸出功能。你可以探索循環播放的角色表情反應、可縮放的向量圖稿，以及精緻的靜態圖像，同時兼顧肌膚、布料、光影與材質等細節。實際可用的格式與控制選項，取決於 Canvas 中所顯示的模型版本與設定。
GPT Image 2.6 圖像範例
探索電影感場景、角色研究與細膩的視覺敘事。
A cinematic, photorealistic movie poster in 16:9 widescreen format.
An elderly man with white hair and a full white beard is driving a vintage, weathered pale blue convertible along a quiet country road. Two young children sit beside him in the car: a brown-haired boy wearing a red T-shirt in the front passenger seat, and a young girl sitting behind him. The elderly man has a thoughtful, slightly serious expression, with both hands on the steering wheel.
The composition closely follows a cinematic road-trip film still: the vintage convertible dominates the foreground, with the windshield framing the three characters. A long, straight rural dirt road stretches into the distance behind them, lined with tall, mature trees on both sides. Vast golden wheat fields extend across the landscape under a pale, softly lit sky.
**Visual style:** Cinematic realism, nostalgic European countryside atmosphere, natural sunlight, warm golden-hour lighting, subtle film grain, muted vintage color grading, earthy golden yellows, dusty greens, faded turquoise, and soft beige tones.
**Lighting and cinematography:** Soft natural daylight, gentle directional sunlight filtering through the trees, realistic shadows, balanced exposure, subtle atmospheric depth, slightly desaturated vintage film aesthetic.
**Camera and composition:** Eye-level frontal camera angle, medium-wide shot, characters positioned centrally, symmetrical road perspective, strong depth of field, visually balanced framing.
High-end cinematic photography, authentic human expressions, realistic skin textures, detailed environment, professional movie poster composition.
No text, no title, no typography, no logos, no credits, no watermarks.
Inside an empty Japanese community gymnasium, an adult brother and sister sit on a bench facing a modest handmade seaside set, pale blue fabric, cardboard waves and white clouds suspended by visible threads, stands and cables remaining on the wooden floor, both figures seen from behind as they quietly watch an artificial but familiar summer, nearly symmetrical two-shot, 16mm spherical film, medium lens, natural soft lighting, faded blue-green-grey palette, coarse film grain, static camera, no camera movement
A cinematic, photorealistic movie poster in 16:9 widescreen format.
A sinister-looking clown stands inside an old-fashioned laundromat, holding an oversized, colorful rainbow spiral lollipop close to his mouth.
The clown wears a dark charcoal-gray hooded sweatshirt with the hood pulled over his head. His face is covered in cracked, weathered white clown makeup, with exaggerated blue diamond-shaped eye makeup, dark smudged eye sockets, a large round red clown nose, and faded red markings above his eyebrows. His makeup appears worn, dirty, and slightly smeared, with realistic skin imperfections and unsettling details.
He has a quiet, disturbing expression, staring directly toward the camera with an intense, melancholic gaze. His head is slightly tilted downward, and he holds the lollipop stick with one hand near his chest. The oversized rainbow lollipop is positioned prominently in the center of the composition, partially covering his mouth.
**Environment:** An empty, slightly run-down American laundromat with rows of white front-loading washing machines extending into the background. The machines create a repeating visual pattern along the right side of the frame. The interior features muted teal walls, fluorescent ceiling lights, and subtle signs of everyday wear.
**Visual style:** Psychological horror thriller, dark cinematic realism, unsettling atmosphere, gritty independent film aesthetic, highly detailed textures, realistic human features, understated horror rather than exaggerated fantasy.
**Lighting and color grading:** Moody, cool-toned fluorescent lighting mixed with subtle warm highlights. Deep shadows around the character, muted teal and desaturated green tones, dark charcoal clothing, and vivid rainbow colors on the lollipop as the main visual contr
A cinematic, photorealistic movie poster in 16:9 widescreen format.
Two men in black business suits, white dress shirts, and black ties are sitting side by side in a retro American diner booth. Both men have blood splattered across their faces, shirts, and suits, suggesting they have just survived a violent incident.
The man on the left has slicked-back dark hair and a slightly bruised, bloodied face. He is leaning back casually against the booth, looking directly toward the camera with a tired, indifferent expression. He holds a fork with a piece of cherry pie in one hand, with a partially eaten pie on a plate in front of him.
The man on the right has dark, slightly messy hair and a bloodied face. He leans comfortably toward his companion, drinking a pink strawberry milkshake through a straw from a tall vintage milkshake glass. His expression is relaxed and almost amused, creating a darkly humorous contrast with the blood covering his clothes.
**Environment:** A nostalgic American roadside diner with orange-red leather booth seating, turquoise countertops, large windows, chrome details, and a warm retro interior. Through the windows, a sunny suburban street with palm trees is visible.
**Foreground details:** A marble-patterned diner table covered with a half-eaten cherry pie, pink milkshake, scattered napkins, drinking glasses, and subtle blood stains, adding to the narrative atmosphere.
**Visual style:** Dark comedy crime thriller, cinematic realism, vintage American movie aesthetic, sophisticated production design, highly detailed textures, authentic human expressions.
**Lighting and color grading:** Warm natural daylight streaming through large diner windows, soft cinematic shadows, rich warm highlights, muted turquoise and coral tones, deep blacks, subtle film grain, and a slightly nostalgic color palette.
**Camera and composition:** Frontal eye-level medium-wide shot, both characters centered and occupying most of the frame, waist-up composition, symmetrical diner booth framing, cinematic depth of field, balanced visual storytelling.
High-end Hollywood movie poster photography, professional cinematic composition, realistic skin textures, natural body proportions, dramatic yet understated storytelling.
No text, no title, no typography, no logos, no credits, no watermarks.
Inside a local Japanese train stopped at a small station, a young Japanese woman sits silently beside the window, her profile overlapping with the reflection of an older woman standing on the platform, cherry trees outside gently washed by bright spring light, several petals attached to the glass, the carriage quiet and nearly empty, medium close shot, natural soft lighting, muted cool blue and pale pink palette, 35mm film grain, shallow depth of field, static camera, no camera movement
GPT Image 2.6 功能特色
動態 GIF
將角色、吉祥物或產品構想轉化為簡短的循環動畫。請描述一個明確的動作、穩定的鏡頭，以及讓主體能夠自然回到起點的初始姿勢。
可縮放的 SVG 輸出
為網站、應用程式與印刷品探索向量圖稿。SVG 輸出功能可透過 Image API 使用；實際支援的格式請查看所選 Canvas 模型提供的選項。
具立體質感的材質
同時引導表面材質與光影表現：無論是編織布料、拉絲金屬、釉面陶瓷或自然膚質。為反射、陰影與紋理提供清晰的視覺參考。
更一致的角色特徵
在微調姿勢、服裝或背景的同時，保留角色的關鍵特徵。請明確說明需要變更的部分，以及哪些臉部、服裝與構圖細節必須維持一致。
如何使用 GPT Image 2.6
先在此構思內容，接著前往 Canvas 選擇設定並開始生成。
- 01
描述圖像內容
在上方提示詞中寫下主體、動作、構圖與光線。若角色身分或既有設計至關重要，請加入參考圖像。
- 02
在 Canvas 中選擇設定
選取 Generate Now 將草稿帶入 Canvas。選擇可用的模型版本、尺寸、畫質與輸出格式，然後檢視預估點數。
- 03
生成與微調
送出生成請求、檢視結果並進行針對性調整。在後續編輯過程中，請持續明確指定已確認的角色與材質細節。
你可以使用 GPT Image 2.6 創作什麼？
角色與電影感場景
打造逼真的人像與故事瞬間，接著圍繞同一個角色探索全新的構圖、服裝與環境。
產品與行銷視覺
精準控制表面材質、光影與視覺識別，創作產品概念圖、專題配圖與多元的行銷視覺變化。
循環動畫與可縮放圖稿
規劃生動的吉祥物循環動畫與適合向量化的設計，並在所選工作流程支援的情況下選擇 GIF 或 SVG 格式。
GPT Image 2.6 創作建議
- 01
為場景設定明確主軸
以主體和單一視覺目標作為開頭，僅加入能強化該目標的構圖、光線與材質細節。
- 02
保持具體的編輯指示
明確指出要變更的部分與要保留的細節。請使用參考圖像來維持特徵一致，而非每次都從頭描述臉部外觀。
- 03
根據輸出格式撰寫提示詞
製作 GIF 時，請描述簡單且可重複的動作；製作 SVG 時，則以清晰的形狀與簡潔的構圖為主。使用前請先檢查細小文字與微小細節。
GPT Image 2.6 計費
捲動查看所有方案比較
| 模型與參數 | Starter 12,000 點數 / 月 | BASIC 30,000 點數 / 月 | Plus 63,000 點數 / 月 | Ultra 120,000 點數 / 月 |
|---|---|---|---|---|
| 圖片生成 | ||||
GPT Image 2.6 Flare 隨生成設定變化 依實際用量結算 | 依實際用量結算 | 依實際用量結算 | 依實際用量結算 | 依實際用量結算 |
GPT Image 2.6 Sunburst 隨生成設定變化 依實際用量結算 | 依實際用量結算 | 依實際用量結算 | 依實際用量結算 | 依實際用量結算 |
以上為預估值，實際費用以最終結算為準。
GPT Image 2.6 替代方案
比較不同模型的工作流程、輸出格式與創作控制度，找出最符合你專案需求的選擇。
GPT Image 2.6
探索更快速的迭代、精細編輯，以及較新的 GIF 與 SVG 輸出工作流程。可依手邊的任務需求選擇 Flare 或 Sunburst。
GPT Image 2.5
搭配 Flare 與 Sunburst 延續既有的圖像生成或編輯工作流程，特別適合現有素材已符合創作方向的情況。
Seedream 5
比較另一種以參考圖為主導的工作流程，適用於產品視覺、角色研究與視覺開發，並提供 Lite 與 Pro 選項。
關於 GPT Image 2.6 的常見問題
GPT Image 2.6 的速度提升了多少？
根據 OpenAI 報告，Images 2.6 的圖像生成速度提升了 50%。實際等待時間會因模型版本、畫質、複雜度、圖像尺寸及服務負載而異。
是否支援 GIF 與 SVG？
官方 Image API 支援動態 GIF 輸出與靜態 SVG 生成。SVG 屬於生成輸出格式，而非圖像編輯輸出格式。Arcloop 中實際可用的格式取決於所選的 Canvas 模型。
我該選擇 Flare 還是 Sunburst？
若要進行快速探索，建議先從 Flare 開始；若需要精細構圖與深度編輯，則可考慮使用 Sunburst。請根據你自己的場景比較兩者的結果，而非假設單一版本適用所有任務。
在編輯過程中可以保留角色特徵嗎？
請使用清晰的參考圖像，並明確指定要保留的臉部特徵、服裝及其他外觀細節。雖然一致性已有所提升，但面對幅度較大的變更時，仍需透過肉眼確認結果。
生成費用如何計算？
在選擇模型與輸出設定後，請於 Canvas 中檢視預估費用。靜態圖像、GIF 與 SVG 請求可能會產生不同的費用；最終收費將以實際生成結果為準。
開始使用 GPT Image 2.6 創作
從詳細提示詞或參考圖開始，將滿意的結果儲存為可重複使用的專案素材。