跳至主要內容
AI 影片模型
Alibaba
Wan

Wan 2.2 A14B Turbo

Wan 2.2 A14B Turbo 在 Kyeo 並非單一工作流程:沒有素材時使用文字生成,上傳一張圖片後改走圖片生成,同時提供一張圖片與一段音訊才會進入語音驅動。三條工作流程的解析度、點數與可見參數不同,選擇前應先確認素材組合。

功能標籤
文字生成、圖片生成、圖片加音訊
點數消耗
標準 40/80;語音按秒計費
提示詞上限
提示詞必填,最多 5,000 個字元
上傳限制
最多 1 張圖;語音另加 1 個音訊;單檔 10 MB
30 秒快速瞭解
  • 三個模型版本會依無素材、單圖或單圖加單音訊動態選擇
  • 580p 僅支援圖片與音訊驅動模式
  • 文字/圖片固定 5 秒為 40/80 credits;語音每輸出秒為 12/18/24 credits
  • 語音工作流程依現行契約傳送預設開啟的 `enable_safety_checker`

熱門搜尋: 如何使用 Wan 2.2 A14B Turbo 模型? · Wan 2.2 三條工作流程分別需要哪些素材? · Wan 2.2 的六檔路由價格如何計算?

示例

Cinematic scene

Main subject: Young Provençal woman, early 20s, natural and sun-kissed everyday appearance (reminiscent of a Pagnol heroine like Manon or Naïs). Wearing a simple, faded en toile light cotton sundress with a subtle floral pattern, a beige linen apron tied around her waist, and worn leather sandals. Thick, wavy dark brown hair loosely tied back with a simple fabric ribbon, with a few stray curls framing her face. Realistic skin texture with light freckles from the sun, absolutely no makeup, warm, expressive, and mischievous personality. Maintain consistent identity, clothing, and appearance throughout the entire video. Location: An authentic, quiet Provençal village nestled at the foot of arid limestone hills during a warm, bright late morning. Narrow cobblestone alleys, old stone houses with faded pastel shutters (pale blue or lavender), small shaded terraces, terracotta pots with geraniums and wild rosemary, laundry lines stretching across the street, a wooden cart, an old stone well, and mature olive trees casting dappled shadows. Quiet, timeless rural atmosphere. No modern elements, no cars, no commercial signs. Visual Style: Ultra-realistic vintage film realism, 1930s-1950s cinema verité. Genuine candid behavior, natural body language, unscripted slice-of-life feeling. Strong historical and regional authenticity, rich real-world textures (stone, dust, rough fabric), and believable human motion. Camera Style: Early 16mm color film aesthetic (like vintage Technicolor or Kodachrome with warm tones). Handheld, immersive, and slightly imperfect framing as if a friend is documenting a summer day. Soft lens texture, gentle organic film grain, warm golden hour light, occasional slight overexposure when stepping into the bright Mediterranean sun, natural lens flares, and soft contrast. No modern stabilization, no digital crispness. Timeline (00:00–00:15) 00:00–00:02Outside the heavy wooden door of an old stone cottage. She sits on a low stone bench, readjusting the ribbon in her hair with both hands raised. A light mistral breeze moves loose strands of her hair. She smiles warmly at the camera with a sparkling, expressive gaze. 00:02–00:04The camera follows her as she walks into a narrow, sun-drenched alley lined with terracotta pots. She spots a stray cicada or a small village dog approaching and crouches down, her dress pooling around her. The framing drifts slightly, capturing the dust dancing in the sunbeams. 00:04–00:06She gently pets the dog (or coaxes a cicada onto a leaf). The camera focus shifts organically between her smiling face and her hands. Bright Mediterranean sunlight flickers through the silver leaves of an olive tree overhead. 00:06–00:08A small patch of dirt yard beside her house. She hangs crisp white linen sheets on a rope clothesline, her arms reaching up as the fabric sways in the wind. The camera catches the bright contrast between the white fabric and the deep blue sky. 00:08–00:10On a quiet, shaded terrace, holding a hand-painted ceramic pitcher. She pours fresh water into a glass, takes a sip, and sits comfortably, watching the quiet village life. She occasionally brushes a stray curl behind her ear. Loose handheld angle with natural camera drift. 00:10–00:12Close side profile. Someone off-camera (an old villager) greets her with a loud, singing voice. She turns her head, raises her hand with a wide, radiant smile, and casually calls out, “Oh, bonjour, Marius !” The camera pans slightly late, capturing her spontaneous laughter. 00:12–00:15Walking slowly down a dusty path towards the hills, holding a small basket of fresh figs. She notices the camera, gives a cheeky, genuine smile, then looks toward the horizon and continues walking. The film reel cuts abruptly to a warm light leak and black, as if the camera ran out of film. Audio: Natural ambient soundscape of the South of France — a loud, continuous chorus of cicadas (les cigales), a light rustling of olive leaves from the Mistral wind, distant clinking of pétanque boules, faint laughter or chatter in a thick Provençal accent from a nearby square, footsteps on dry earth and stone. No modern music. No artificial sound design. Pure, immersive atmosphere. Goal: Authentic Provençal village life captured like a lost piece of poetic realist cinema — candid, sun-drenched, imperfect, warm, and deeply nostalgic.

Cinematic scene

Create a 15-second photorealistic Korean shampoo commercial using @image1 as the exact reference for the female model, shampoo bottle, bathroom, hairstyle, and visual style. A young Korean woman looks at her long hair in a bright white bathroom and notices that it appears dry and lacks softness. She uses Lilac Bloom Shampoo, dispenses the clear gel into her palm, then washes her hair with rich soft foam. Show elegant close-ups of the shampoo bottle, pump, foam, wet hair, and clean water. After drying, her long black hair looks smooth, healthy, glossy, and moves naturally in a gentle breeze. End with the same woman smiling at the camera while holding the shampoo bottle beside her face and saying in Korean: 「릴락 블룸으로, 오늘도 부드럽게 빛나.」

Cinematic scene

A 15-second 2D flat motion-graphics commercial in a clean, rich chocolate-drink style. Soft warm beige / pale cream paper-textured background. Limited palette: deep cocoa brown, creamy off-white, milk-chocolate amber, gold sparkle accents, and thin warm-cream highlight lines. Smooth, weighty liquid animation, glossy pours, and soft squash-and-stretch. No photorealism — bold graphic shapes with slight paper grain. 0.0–1.2s Scattered glossy dark-brown chocolate droplets of different sizes float and drift slowly across the empty cream background, each with a small cream highlight. 1.2–2.5s The droplets gather and pour downward in a thick, ribbon-like stream, forming a white-outlined short tumbler glass. Rich brown liquid fills it from the top and overflows in a rounded, velvety splash. 2.5–4.0s A second graphic can tilts in from the top-right and pours a continuous chocolate stream into the glass. The liquid fills completely; cream-colored circular swirls rise inside. Small gold sparkles appear around the rim and splash. 4.0–6.0s The finished drink: short white-outlined glass of glossy chocolate, condensation dots as simple white circles, “Cokoa” lettering on the glass. A standing can appears beside it. Graphic chocolate chunks and a cocoa-powder pile slide in at the bottom-left. Blue-cream curved splash lines and gold four-pointed stars pop around the scene. The camera eases slightly closer; the pour feels thick, creamy, and alive. 6.0–8.5s The drink and splash dissolve into large, bold, condensed cream typography that reads THE PERFECT POUR. The letters first appear as if they are filled with the same flowing chocolate and cream swirls, then the liquid drains away leaving solid cream lettering sitting on a wide, organic dark-brown splash / wave shape that looks like overflowing chocolate. 8.5–15s The title stays locked on screen. Small brown droplets continue to bounce on and around the letters. Supporting text types on underneath in a clean condensed cream font: first CREAMY. then expanding to the short product line Indulge in the rich and creamy goodness of Cokoa. A tiny logo lockup appears at the bottom center: the word Cokoa plus a miniature standing can. Subtle swirl, droplet, and sparkle animation continues until the end. Mood: rich, satisfying, slightly luxurious everyday chocolate treat. Timing should feel snappy and weighty — liquids have body, cream is soft, pours have momentum, text has a juicy “splash” reveal. Keep everything graphic, flat, and highly readable.

Cinematic scene

Create a second ultra-realistic luxury beach beauty commercial in vertical 9:16 format. Use the same adult female model, same face, hairstyle, skin tone, and same colorful beach outfit from the reference image throughout the entire video. Keep the product design consistent in every shot: SunKiss Beach Glow Shimmer Mist, turquoise-to-coral gradient bottle with a gold spray top. Scene 1 — Sand Formation Open with a cinematic macro shot of smooth golden beach sand beside the ocean. The exact shape and silhouette of the SunKiss bottle is naturally formed and embossed into the sand, including the bottle outline and cap shape. Warm sunlight, tiny grains of sand visible, soft ocean ambience in the background. Scene 2 — Ocean Transformation A gentle turquoise ocean wave slowly moves toward the sand-shaped bottle. In slow motion, clear foamy seawater flows directly over the sand impression. As the water passes across it, create a seamless magical transformation: the sand-made bottle gradually becomes the real premium SunKiss Beach Glow Shimmer Mist bottle. Start from one end and transform progressively until the complete physical bottle appears. Show water droplets sliding across the turquoise and coral bottle, golden cap reflecting sunlight, extremely realistic product textures and reflections. Scene 3 — Model Enters Cut to a wide cinematic beach shot. The same model walks naturally toward the product with the ocean and palm trees in the background. Soft breeze moves her hair and colorful beach outfit naturally. She notices the bottle lying near the shoreline, bends down naturally, picks it up, looks at it with a soft confident smile, and turns it slightly toward the camera so the product is clearly visible. Scene 4 — Product Application Show several premium beauty-ad close-ups. First, she sprays a light mist onto her shoulder and arm. Then show a close-up of her gently spreading the product over her arm. Next, show her applying a small amount to her legs. The product creates a natural hydrated summer glow with subtle shimmer, realistic healthy-looking skin and sunlight reflections. No exaggerated glitter. Include macro slow-motion shots of the fine mist catching the sunlight. Scene 5 — Lifestyle Promotion Create quick elegant promotional shots from different camera angles: - model holding the product beside her face - close-up of the bottle in her hand with ocean bokeh behind it - model walking along the shoreline holding the bottle - bottle standing upright in wet sand while a soft wave passes behind it - close-up of golden sunlight reflecting from the bottle - model smiling naturally after applying the product Keep movements elegant, relaxed and premium, like a professional summer skincare campaign. Final Hero Shot Finish with the SunKiss bottle standing beautifully on wet golden sand near the shoreline. A gentle wave stops just around the bottle while golden sunset light reflects through the transparent packaging. Camera performs a slow cinematic push-in toward the product. Audio: soothing ocean waves, gentle summer ambience, soft tropical instrumental music, subtle water splash sounds, fine mist spray sound, no loud music. Visual style: ultra-photorealistic, premium summer beauty commercial, cinematic beach photography, natural golden sunlight, turquoise ocean, realistic water physics, smooth transitions, shallow depth of field, macro product photography, realistic skin texture, professional advertising cinematography, 4K, no CGI appearance, no distorted hands, no changing product design, no changing model identity.

Cinematic scene

Create an ultra-realistic Get Ready With Me cinematic lifestyle video where the girl in @/image1 is preparing to leave home to watch a daytime football match. The video should feel like a luxury TikTok / Instagram Reel vlog mixed with cinematic commercial quality. Opening scene: soft natural daylight entering a modern apartment bedroom. The girl films herself casually in mirror-style shots while smiling naturally. Cozy clean aesthetic, warm sunlight, realistic home environment. Sequence includes: choosing a football jersey outfit from the closet, close-up shots of her applying subtle makeup and lip gloss, fixing her hair naturally in the mirror, putting on jewelry and perfume, trying on the football jersey in @/image2 with oversized casual sporty styling, cinematic close-ups of sneakers, jersey details, accessories, and phone, natural candid moments laughing at the camera, drinking iced coffee while getting ready, packing essentials into a small shoulder bag. The pacing should feel trendy and smooth with realistic influencer-style transitions. Handheld vlog camera mixed with cinematic tracking shots. Then show her leaving the apartment building confidently during bright daytime weather. Hair moving naturally in the wind, sunlight hitting her face beautifully. She walks through busy streets toward the football stadium while excited fans in jerseys pass by around her. Final sequence: entering the packed football stadium, walking through the crowd smiling excitedly, sitting in the stands watching the match, close-up reaction shots cheering and laughing, energetic daytime football atmosphere with chanting fans and crowd noise. Ultra realistic lifestyle cinematography, premium influencer aesthetic, realistic skin texture, natural movement, shallow depth of field, bright daylight, smooth transitions, authentic social media reel energy, cinematic commercial quality, 4k realism.

Cinematic scene

A surreal vertical night scene of a lonely man standing in a vast dark desert under a star filled sky. In front of him is a gigantic freestanding glowing door, many times taller than a human, slightly open, revealing a bright orange light inside. Around the door, sand softly glows under blue moonlight. The sky is rich with stars and faint nebula clouds. Strong visual contrast between the cool midnight blue landscape and the warm fiery interior light spilling from the door. Quiet, symbolic, cinematic, mysterious, highly detailed, surreal scale, emotional solitude. Negative: no text, no logo, no watermark

一句話瞭解

這個模型表現如何?

一句話瞭解 — 580p 僅支援圖片與音訊驅動模式. 480p / 720p: 40 / 80 點數. 語音生成影片: 480p / 580p / 720p = 12 / 18 / 24 點數 × (num_frames ÷ frames_per_second).

它擅長什麼
依輸入素材選擇文字生成影片、圖片生成影片或語音驅動影片。三種模式使用各自的輸入和參數。
適合哪些使用者
適合分別驗證純文字生成、單圖動畫,或以單圖搭配單一音訊驅動人物的影片任務;三條路徑應各自評估,不宜把其中一條的表現推論到另外兩條。
為什麼要在 Kyeo AI 使用
Kyeo 依素材組合自動選擇三條工作流程,並在送出前顯示標準固定 5 秒價格與語音每秒費率;每條路徑仍須分別驗收。

關鍵參數總覽

協助你快速評估此模型是否符合你的業務情境。

模型類別
AI 影片模型
廠商
Alibaba
模型家族
Wan
串接方式
目前站內串接
功能標籤
Wan — 非同步任務
執行方式
非同步任務
提示詞上限
Kyeo 請求:提示詞不可空白,最多 5,000 個字元
上傳限制
JPEG/PNG/WebP;語音另需 1 個音訊;單檔 10 MB

其他名稱

同一個模型在官方文件、社群討論與 API 串接層可能有不同稱呼,這裡幫你統一整理。

Wan — Wan 2.2 哪個工作流程開放 enable_safety_checker?
Wan — Wan 2.2 只上傳音訊會發生什麼事?
Wan — Wan 2.2 與 Wan 2.6 應如何選擇?

使用者常見問題

從實際任務、輸入條件和結果要求出發,整理選擇模型時最需要確認的問題。

如何使用 Wan 2.2 A14B Turbo 模型?
Wan 2.2 三條工作流程分別需要哪些素材?
Wan 2.2 的六檔路由價格如何計算?
580p 僅支援圖片與音訊驅動模式
Wan 2.2 圖片路由接受哪些格式?
Wan 2.2 語音路由需要圖片和音訊嗎?
Wan 2.2 的 num_frames 為何必須是 4 的倍數?
Wan 2.2 哪個工作流程開放 enable_safety_checker?
Wan 2.2 只上傳音訊會發生什麼事?
Wan 2.2 與 Wan 2.6 應如何選擇?

選擇指南

還在猶豫該用哪個模型嗎?先查看這幾個關鍵決策點。

1
先依素材確認工作流程

沒有素材時只測文字工作流程;一張圖片且沒有音訊時測圖片工作流程;語音工作流程必須同時提供一張圖片和一段音訊。

2
再核對解析度與價格

580p 僅支援圖片與音訊驅動模式

3
送出前檢查進階參數

語音模式的 `num_frames` 使用 40~120 之間的 4 的倍數,影格率使用 4~60。

使用者常搜尋的比較

把常見候選放在同一任務下比較,協助你確認輸入方式、控制項和成本差異。

Wan 2.2 與 Wan 2.6 的素材工作流程有何不同?
Wan 2.2 vs Wan 2.6:解析度與點數
Wan 2.2 與 Wan 2.5 Video 的參數有何不同?
Wan 2.2 vs Wan 2.5 Video:輸入與價格
Wan 2.2 語音路由與 Infinitalk 有何不同?
Wan 2.2 vs Infinitalk:圖片、音訊與提示詞

模型比較表

還在猶豫該選哪一個嗎?這張表協助你快速看清目前模型與替代方案的核心差異。

素材觸發路由
Wan 2.2 A14B Turbo
無素材、單圖、單圖加單音訊
Wan 2.6
無素材、單圖或單影片
Infinitalk
固定要求單圖、單音訊與提示詞
目前點數
Wan 2.2 A14B Turbo
文字/圖片固定 5 秒為 40/80;語音每秒 12/18/24
Wan 2.6
70-315 點數,依解析度與時長
Wan 2.5 Video
60-200 點數,依解析度與時長
解析度選項
Wan 2.2 A14B Turbo
580p 僅支援圖片與音訊驅動模式
Wan 2.6
720p、1080p
Wan 2.5 Video
720p、1080p
可見控制項
Wan 2.2 A14B Turbo
文字/圖片開放標準控制;語音另有影格、推論與安全檢查控制
Wan 2.6
解析度與時長 2 項
Infinitalk
解析度與 seed 2 項
目前驗證邊界
Wan 2.2 A14B Turbo
音訊須搭配圖片;影格數須為 4 的倍數
Wan 2.6
影片工作流程不使用圖片或音訊
Infinitalk
圖片與音訊固定同時必填

實際使用體驗

結合公開資料、目前站內邊界和典型任務,說明模型適合與不適合的情況。

目前公開資料介面工作流程名稱不等於 Alibaba 直連型號

Alibaba 官方 Wan 2.2 A14B 發布資料涵蓋文字與圖片模型;Alibaba Cloud 目前的 `wan2.2-s2v` 還有獨立圖片檢測步驟。目前公開資料介面的 A14B 語音工作流程只依目前公開資料介面表單與 Kyeo 傳送內容陳述。

產品宣傳不保證特定結果

720p、24 fps、快速生成、穩定動作與影音同步都不是 Kyeo 的結果保證;應以實際任務輸出為準。

功能詳解

文字與圖片模式

文字模式可設定提示詞、解析度、長寬比、提示詞擴展、加速和選填的隨機種子;圖片模式改為上傳一張參考圖,不再另外選擇長寬比,其餘可見控制項維持一致。安全檢查固定啟用。

語音驅動模式

語音模式需要提示詞、一張圖片和一段音訊,並提供影格數、影格率、解析度、推理步數、引導強度與偏移量控制;負面提示詞和隨機種子可選,安全檢查預設啟用。

安全欄位邊界

所有模式都會啟用安全檢查,但自動檢查不能取代人工內容安全與素材權利審核。

適合做什麼

純文字鏡頭測試

只用提示詞比較 16:9 與 9:16、480p 與 720p;結果須人工檢查主體、動作、文字與畫面連續性。

單圖動態化

上傳一張 JPEG、PNG 或 WebP 圖片,並為動作和鏡頭寫清楚提示詞。

圖片加音訊驅動

同時上傳一張圖片與一段支援的音訊後,依影格數和影格率測試口型或動作;正式發布前必須人工核對同步、身分、音訊和素材權利。

提示詞技巧

先描述可見動作

說明主體、場景、動作與鏡頭移動,避免把速度、穩定性或最終品質寫成提示詞中的確定結果。

圖片工作流程沒有比例控制

圖片模式不送出 aspect_ratio;先確認原圖構圖,再說明希望保留或改變的主體動作,並檢查輸出裁切與邊緣。

分開處理音訊與畫面指示

音訊提供聲音輸入,提示詞仍用於描述人物狀態與畫面動作。negative_prompt、guidance_scale 和 shift 只是請求控制項,不等於同步或品質保證。

選擇它的理由

同一頁面可依素材切換三條明確的目前公開資料介面模型版本。
文字與圖片路由提供比例、提示詞擴寫、seed 和 acceleration 的精簡控制集合。
全部工作流程預設啟用安全檢查,語音模式也維持開啟
六檔 Kyeo 工作流程價格與對應解析度可在送出前直接核對。

使用前須知

使用 Wan 時,建議先依所選參數生成一個代表性樣本,確認效果符合目標後再擴大批次。
580p 僅支援圖片與音訊驅動模式
語音模式的 `num_frames` 必須是 40~120 之間的 4 的倍數。
Turbo、24 fps、動作穩定、影音同步、畫質、安全過濾與生成時間會受輸入和目前公開資料狀態影響,不構成 Kyeo 的結果保證。

可調整參數

協助你快速評估此模型是否符合你的業務情境。

解析度
選填
參數類型: select
預設值:720p
480p
580p
720p
畫面比例(僅文字路由)
選填
參數類型: select
預設值:16:9
16:9
9:16
提示詞擴寫(文字/圖片)
選填
參數類型: select
預設值:false
關閉
開啟
隨機種子
選填
參數類型: number
範圍 0-2147483647
加速層級(文字/圖片)
選填
參數類型: select
預設值:none
none
regular
影片影格數(語音)
選填
參數類型: number
預設值:80
範圍 40-120,必須為 4 的倍數
每秒影格數(語音)
選填
參數類型: number
預設值:16
範圍 4-60
負面提示詞(語音)
選填
參數類型: text
推論步數(語音)
選填
參數類型: number
預設值:27
範圍 2-40
提示詞引導強度(語音)
選填
參數類型: number
預設值:3.5
範圍 1-10
偏移值(語音)
選填
參數類型: number
預設值:5
範圍 1-10
安全檢查(僅語音)
選填
參數類型: select
預設值:true
關閉
開啟

點數消耗

標準 40/80;語音 12/18/24 每秒

580p 僅支援圖片與音訊驅動模式. 480p / 720p: 40 / 80 點數. 語音生成影片: 480p / 580p / 720p = 12 / 18 / 24 點數 × (num_frames ÷ frames_per_second).

省點數小技巧

批次生成圖片或影片前,建議先用同一組素材,在目前模型與同類候選模型之間進行 A/B 測試;確認效果後再批次生成,避免浪費點數。

常見問題

同類模型推薦

還沒決定該用哪個嗎?順便比較一下這幾個同類候選模型。

Wan 2.6

不附媒體時使用文字生成影片,恰好 1 張圖片時使用圖片生成影片,1–3 個影片時使用影片編輯;圖片與影片不可混用。 預設為 1080p、5 秒、鏡頭結構關閉、內容過濾開啟。

AI 影片模型
70–315 點數/次

Wan 2.5 Video

瞭解 Wan 2.5 Video 在 Kyeo AI 的文字與單圖影片流程、5/10 秒、720p/1080p、提示詞差異及 60–200 點數計費。 Kyeo 依解析度與長度精確計費:720p 5 秒 60 點數、720p 10 秒 120 點數、1080p 5 秒 100 點數、1080p 10 秒 200 點數;預設 720p 5 秒為 60 點數。Wan 2.5 的 60/100/120/200 點數如何計算? — 瞭解 Wan 2.5 Video 在 Kyeo AI 的文字與單圖影片流程、5/10 秒、720p/1080p、提示詞差異及 60–200 點數計費。

AI 影片模型
每次 60/100/120/200 點數

Infinitalk

Infinitalk 在 Kyeo 使用 1 張圖片、1 個不超過 15 秒的音訊檔案與必填提示詞,生成虛擬人物播報影片。可選 480p 或 720p,也能填寫 `seed`;費用依音訊秒數與解析度計算,不是固定的單次價格。

AI 影片模型
480p 每秒 3 點數;720p 每秒 12 點數

資訊來源

頁面內容依據廠商資料、模型系列說明與 Kyeo AI 平台串接設定整理。

來源說明

瞭解 Wan 2.2 A14B Turbo 的文字、單圖與語音驅動流程、標準固定 5 秒價格、語音每秒費率及素材契約。 580p 僅支援圖片與音訊驅動模式. 480p / 720p: 40 / 80 點數. 語音生成影片: 480p / 580p / 720p = 12 / 18 / 24 點數 × (num_frames ÷ frames_per_second).

最後更新: 2026-09-15
Alibaba Cloud:Wan2.2 A14B 發布說明
官方
查看原始來源
Alibaba Cloud:wan2.2-s2v 直連說明
官方
查看原始來源