跳至主要內容
Kyeo AI
AI 影片模型
Alibaba
Wan

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。

使用此模型查看點數費用回到模型廣場
功能標籤
文字 / 1 張圖片 / 1–3 個影片
點數消耗
70–315 點數／次
提示詞上限
1–5,000 / 2–5,000
上傳限制
1 JPG/PNG/WebP · 1–3 MP4/MOV · 10 MB
30 秒快速瞭解
  • Wan 2.6 — 素材驅動三工作流程: 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。
  • Wan 2.6 — 有效時長: 文字與圖片輸入支援 5、10 或 15 秒；影片輸入只支援 5 或 10 秒。
  • Wan 2.6 — 六種 Kyeo 點數價格: 720p 的 5/10/15 秒分別為 70/140/210 點，1080p 分別為 105/210/315 點。影片輸入不提供 15 秒。
  • Wan 2.6 — 可見控制與素材: 只接受 1 張至少 256×256 且不超過 10 MB 的 JPG/PNG/WebP 圖片，或 1–3 個每個不超過 10 MB 的 MP4/MOV 影片。開始任務前，每個檔案都必須通過格式、大小、尺寸與簽章中繼資料驗證。

熱門搜尋: Wan 2.6 怎麼用 · Wan 2.6 六種價格 · Wan 2.6 如何選擇文字工作流程

示例

Cinematic scene

Create a 15-second vertical 9:16 kinetic typography product video for any Nike shoe using @ Image1 as the main reference. Treat @ Image1 as the shoe’s full visual DNA. Analyze the shoe’s silhouette, proportions, colorway, materials, sole design, cushioning, textures, overlays, laces, and overall performance/lifestyle attitude, then make every typography choice feel physically derived from the shoe. The shoe must cause the typography to move. Do not place random animated text over the product. Use premium editorial kinetic typography with bold sans-serif letterforms. Let the type weight, width, spacing, distortion, color, texture, and motion adapt to the shoe’s identity. If the shoe feels fast, make the type stretched, condensed, and forward-driven. If it feels cushioned, make the type compress and rebound. If it feels chunky or basketball-oriented, make the type heavy, impactful, and explosive. If it feels futuristic, make the type sleek, dimensional, and experimental. Structure: 0–3s strong hook with macro product reveal and oversized typography interacting instantly with the shoe. 3–6s reveal the shoe details through fast editorial cuts while the typography forms from shoe lines, textures, and motion. 6–10s hero motion sequence where the shoe rotates, lands, accelerates, or cuts through space while words wrap around it, split apart, get sliced, compressed, stretched, or revealed through depth. 10–13s show 2–4 short kinetic words driven by the shoe’s character, such as SPEED, AIR, GRIP, FLY, IMPACT, MOVE, or ELEVATE. Use the model name only if it is visible or clearly recognizable from the reference. Never invent claims or tech not supported by the shoe. 13–15s end with a clean iconic hero lockup: the shoe centered, typography resolved, subtle Nike branding, premium finish. Visual style: ultra-clean, bold, high-contrast, premium Nike ad energy, photoreal product detail, sharp lighting, dynamic shadows, controlled motion blur, graphic backgrounds derived from the shoe palette, fast cuts, match cuts, typography-driven transitions, crisp readable text, no clutter, no extra objects unless they support the typography interaction. Camera language: macro push-ins, lateral tracking, orbit shots, top-down impact views, side-profile hero angles, and aggressive close detail shots. Keep the pacing energetic, stylish, modern, and satisfying.

Cinematic scene

MARC JACOBS — "SUMMER GLOSS" Lipstick Commercial **ACTIVE REFERENCES** **MODEL REFERENCE** Exact same woman as the bag campaign. Identical facial identity, hairstyle, body proportions, confident editorial expression, posture, walking style. Natural makeup base with the hero lipstick applied (rich, vibrant pink-rose gloss). She wears a soft, flowing babypink dress (lightweight silk-cotton blend, slightly billowy sleeves, midi length, elegant yet effortless summer style). Maintain identical identity throughout. No drift. **@Image 1 — MARC JACOBS LIPSTICK** Hero product reference. Sleek metallic tube in warm rose-gold finish with signature MJ engraving. Creamy, high-shine gloss in a fresh pink-rose shade that catches light beautifully. Maintain exact tube design, color, logo placement, and glossy bullet texture. Never redesign. **@Image 2 — LIP CLOSE-UP & APPLICATION** Reference for extreme beauty macros: perfectly applied glossy lips, micro texture of the gloss, light refraction on the lips, subtle lip movement when speaking or smiling faintly. **LOCATION** Busy Manhattan intersection. Warm late-afternoon golden sunlight. Historic brownstone buildings. Yellow taxis. Crosswalk. Light traffic. Editorial New York energy. Same architecture and atmosphere as the bag film. **FORMAT** Single continuous luxury fashion film. 45 seconds. Horizontal 16:9. Museum-quality realism. No AI look. No CGI. Photoreal filmic beauty. **DIRECTOR'S NOTES** This is not just a lipstick ad — it’s about radiant confidence that never fades. She moves through the fast city on her own terms. The gloss stays impossibly luminous and glossy under direct sunlight, even as she walks, speaks, and smiles. The film celebrates the product’s staying power and the effortless glow it gives her. Alternate constantly between: • Movement & confidence • Beauty close-ups • Product texture & application • Sunlit glamour Every 3–4 seconds the lipstick (or her lips) returns as the hero. Every cut feels expensive. Think Marc Jacobs beauty campaign meets Vogue beauty film. **COLOR PALETTE** Warm golden sunlight, soft babypink tones, rose-gold accents, rich skin tones, honey highlights. Kodak Vision3 film look. Subtle halation. Soft film grain. Natural skin. No teal-orange. **CAMERA** Organic handheld + editorial dolly. Short slider moves. Gentle whip pans. Macro beauty photography. Natural lens breathing. Occasional foreground wipes with taxis/pedestrians. **LENSES** 100mm Macro | 85mm Portrait | 50mm | 35mm Street | 24mm Tracking. All behave like real cinema glass with beautiful falloff. --- **SEGMENT 1 (0.0–6s)** Ground level 24mm tracking shot. Camera starts at knee/hip height. Flowing babypink dress hem, strappy sandals, and her confident stride fill the frame. A yellow taxi whips across foreground. At 3 seconds — cut to 100mm macro: she pulls the MARC JACOBS lipstick from her bag, twists it up. Golden sunlight hits the rose-gold tube and the glossy bullet. Extreme texture detail. She applies it smoothly in one fluid motion (mirrorless, on the move). Lips transform into high-shine perfection. **SEGMENT 2 (6–14s)** 50mm medium portrait. Camera tracks backward as she walks calmly across the crosswalk. Babypink dress flows gently in the breeze. Hair moves naturally. Effortless, confident expression (soft, knowing half-smile). Golden sunlight kisses her face. Her lips remain mirror-glossy, catching every ray of light. A taxi blurs behind her. **CUT to fast 0.3-second macro montage** (lips + product): - Gloss catching sunlight, no fading - Creamy texture close-up - Lip movement as she gently presses lips together - Light refracting on the high-shine surface **SEGMENT 3 (14–23s)** 35mm three-quarter tracking shot, matching her walking speed. Side angle. Dress sleeve billows. She turns her head slightly toward the sun, lips prominently featured. Foreground pedestrians and a passing bicycle create natural wipes. **CUT to 100mm macro:** Fingers holding the lipstick tube. She twists it down slowly. Rose-gold metal reflects warm sunlight. Subtle embossed logo visible. Natural reflections only. **SEGMENT 4 (23–32s)** 85mm hero portrait. She pauses at the corner, golden sunlight wrapping her face and shoulders. Babypink dress looks ethereal. She turns slightly toward the city horizon, then glances toward camera with quiet confidence. Lips still impossibly glossy and fresh. Slow push-in on her face. **Voice-over begins** (warm, intimate, luxurious female voice): “In the heat of the city… one gloss refuses to fade.” **SEGMENT 5 (32–40s)** Dynamic montage: - Walking again (medium shot) - Quick macro of lips in direct sun — gloss still perfect - She laughs softly (natural micro-expression) — gloss stays vibrant - Hand bringing lipstick tube back into frame for a final beauty shot - Final macro of lips + tube side-by-side, both glowing in golden hour light **Voice-over continues:** “Marc Jacobs Summer Gloss. Stays luminous. Stays you.” **SEGMENT 6 (40–45s)** Hero product lock. 100mm macro. Lipstick tube fills frame with perfect sunlight raking across the rose-gold metal and glossy bullet. Hold on the texture. Soft fade to black with elegant MJ logo and product name. **PERFORMANCE** Luxury editorial restraint. Natural walking rhythm. Relaxed shoulders. Real micro-expressions. Confident but never over-posed. Slow breathing. The gloss is the star — always pristine. **PRODUCT LOCKS** Lipstick tube and color must be identical to reference. Extreme gloss retention in sunlight is the hero claim. No color fade. No smudging. Natural lip movement with perfect shine maintained at all times. **PHYSICS & LIGHTING** Dress flows realistically in city wind. Hair reacts naturally. Golden hour sunlight with soft building bounce. Natural skin tones and highlight roll-off. Lip gloss has visible, believable shine and light refraction. **AUDIO** Premium luxury electronic track with warm analog bass and subtle percussion. Layered diegetic city sounds (taxis, footsteps, soft wind). Delicate gloss application sound, fabric rustle, and lipstick cap clicks during macros. Voice-over is intimate and aspirational.

Cinematic scene

Create a 45-second 3D Pixar-style animated commercial for GLOWDROP – 24-Hour Liquid Blush in a vertical smartphone format. Use bright, colorful visuals with expressive cartoon characters, smooth exaggerated animation, cinematic lighting, and dynamic camera movements. The story takes place in a cozy apartment bathroom during a warm morning. Zara, a young woman in pajamas and a satin robe, panics after seeing a notification that her first date is in 20 minutes. Her regular powder blush fades instantly, leaving her frustrated, while her roommate's fluffy white Pomeranian, Milo, silently watches with hilarious judgmental reactions. After desperately searching through makeup, Zara discovers the glowing GLOWDROP bottle. She applies two dots, instantly revealing a radiant, long-lasting flush that transforms her confidence. A fast-paced montage shows her finishing her look, confidently heading out, and impressing her waiting date, while Milo proudly follows along. End with a clean hero product shot of the rotating GLOWDROP bottle against a soft pink background with the tagline: "GLOWDROP — Blush That Doesn't Quit." Keep the tone playful, relatable, and comedy-driven, featuring expressive character animation, internet-style humor, glowing beauty visuals, and Milo's silent reactions as the scene-stealing sidekick.

Cinematic scene

A cinematic 30-second summer short film featuring the same young woman throughout the entire video, set in a peaceful American countryside during golden hour. She has shoulder-length dark hair, natural makeup, and a soft youthful appearance. The video opens with a close-up side profile of her standing in a wide green meadow under a dramatic blue sky with soft clouds. Gentle sunlight falls across her face as the wind moves her hair naturally. Cut to a medium shot of her walking slowly through the field, turning her head back toward the camera with a calm, slightly nostalgic expression. Keep her facial identity, hairstyle, body proportions, and outfit consistent. Transition to an intimate overhead and side-angle shot of her lying in tall green grass, looking upward at the sky. The grass moves softly in the breeze, creating a relaxed summer atmosphere. Then move to a rustic countryside location with a classic red wooden barn and fence in the background. She casually leans against the fence, playing with her hair, with warm sunlight creating natural lens flare and soft rim light. For the final scene, she sits alone on outdoor bleachers during sunset, wearing a white tank top and relaxed sporty shorts, holding a dark glass bottle near her face. Warm backlight shines through her hair as she closes her eyes and enjoys the moment. End with a dreamy soft-focus shot and subtle cinematic fade, evoking the feeling of a nostalgic summer memory. Realistic photography, cinematic natural lighting, golden hour sunlight, soft film grain, shallow depth of field, gentle handheld camera movement, natural body movement, realistic skin texture, subtle wind, warm nostalgic summer mood, consistent character identity, smooth scene transitions, emotional visual storytelling, indie coming-of-age film aesthetic, high-end music video cinematography, 9:16 vertical format.

Cinematic scene

10-second photorealistic cinematic comedy video of a giant realistic pinkish-brown octopus working as a car-wash attendant at a busy roadside car wash. A silver 1990s-style sedan is covered in thick white soap suds. The octopus is absurdly energetic, using all eight tentacles simultaneously to scrub and wash the car with yellow sponges and soapy water. A confused male driver sits inside the car watching the bizarre situation unfold. Busy traffic, urban freeway environment, warm late-afternoon sunlight, realistic reflections, water droplets and foam, believable octopus anatomy and natural tentacle movement. The humour comes from treating the giant octopus completely seriously as a professional car washer. Timeline Breakdown: 0–2s: Wide frontal establishing shot. A silver sedan sits at the car wash, completely covered in soap. The giant octopus stands beside it, already holding several yellow sponges with different tentacles. The driver looks through the windshield in disbelief. Camera slowly pushes toward the car. 2–5s: The octopus suddenly goes into overdrive. All eight tentacles rapidly scrub different parts of the vehicle simultaneously windshield, roof, hood and side windows. Thick foam flies everywhere. Water sprays toward the camera. The driver looks increasingly shocked. 5-7.5s: Fast, energetic cinematic movement around the front corner of the car. Tentacles whip across the windshield and roof in exaggerated but physically believable motions. Soap and water cascade down the glass. One tentacle vigorously scrubs the windshield directly in front of the bewildered driver. 7.5-9s: The octopus suddenly stops and proudly steps back. The car is now extremely clean and sparkling. The driver gives the octopus a confused look, then starts driving forward. 9-10s: The silver car drives away. The octopus remains beside the wash station holding its sponges, watching the car leave. End on a slightly wide comedic shot as traffic passes behind it. Visual Style: Photorealistic live-action, cinematic comedy, realistic practical-effects appearance, natural skin and suction-cup detail, physically accurate water and soap physics, realistic vehicle reflections, shallow depth of field, subtle handheld camera movement, warm natural sunlight, high-detail 4K cinematic image quality. Camera: Professional cinema camera, 35mm lens, realistic motion blur, natural exposure, occasional water droplets hitting the lens, smooth cinematic tracking combined with brief energetic handheld movement. Important: Keep the same octopus, car, driver, environment and lighting consistent throughout. No cartoon appearance, no animation, no extra limbs, no deformed tentacles, no morphing, no text, no subtitles, no logos. The octopus must remain anatomically believable despite the absurd comedic premise.

Cinematic scene

[0:00–0:02] Title Card: Clean minimal text animation: DESERT REVERIE fades in centered, thin serif white font, subtle letter spacing expand animation, over a blurred desert background (soft bokeh of red mesas and blue sky). Text holds for 1.5s then fades out smoothly. [0:02–0:05] Establishing Shot: Wide static shot, Monument Valley desert at midday. A vintage teal 1960s muscle car sits parked on a dirt road, red sandstone mesas in the background, clear blue sky. A man in a fringed jacket and cowboy boots reclines on the hood, hands behind his head, completely relaxed. Heat haze shimmers faintly above the ground. [0:05–0:09] Rising Action: Low angle GoPro shot from near the front bumper, slowly tilting upward. A powerful brown horse leaps into frame from the left, mid air, legs tucked, mane and tail whipping with motion. Camera holds steady as the horse arcs gracefully over the car and the reclining man, muscles rippling naturally, correct anatomy maintained (four legs, proper joint bends, no warping). [0:09–0:13] Landing & Dust: Camera transitions to a slow dolly in from the front of the car. Horse lands just behind the vehicle, front hooves touching down first, kicking up a dramatic cascade of dust and sand in realistic particle simulation. Man on the hood remains still and unfazed, jacket fringe fluttering slightly in the wind. Golden desert light, hard natural shadows. [0:13–0:15] Final Hold: Slow motion ramp (0.5x) on the settling dust cloud, mesas sharp in the background, sky calm and clear. Gentle fade to black. Technical requirements: Horse: exactly 4 legs, correct joint articulation, no limb warping, no extra hooves. Man: correct hand/finger anatomy (5 fingers each), no facial morphing. Car body lines and wheels remain stable, no distortion. Consistent dust/sand particle physics, no flicker artifacts. No watermarks, logos, or text overlays except the title card. 16:9 aspect ratio throughout, no letterboxing Color grading: warm desert tones, teal car accent, natural sky gradient, cinematic contrast.

一句話瞭解

這個模型表現如何？

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。 文字與圖片輸入支援 5、10 或 15 秒；影片輸入只支援 5 或 10 秒。

它擅長什麼
Wan 2.6 — 功能標籤: 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。
適合哪些使用者
Wan 2.6 — 素材與安全邊界: 只接受 1 張至少 256×256 且不超過 10 MB 的 JPG/PNG/WebP 圖片，或 1–3 個每個不超過 10 MB 的 MP4/MOV 影片。開始任務前，每個檔案都必須通過格式、大小、尺寸與簽章中繼資料驗證。
為什麼要在 Kyeo AI 使用
Wan 2.6 — 影片長度: 文字與圖片輸入支援 5、10 或 15 秒；影片輸入只支援 5 或 10 秒。 鏡頭結構預設關閉，內容過濾預設開啟；兩個控制項都會顯示並明確送出。

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
執行方式
非同步任務
提示詞上限
文字：1–5,000 個字元；圖片／影片：2–5,000 個字元
上傳限制
1 張 JPG/PNG/WebP 圖片（至少 256×256、最大 10 MB）或 1–3 個 MP4/MOV 影片（每個最大 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。

Alibaba Wan 2.6
Wan 2.6

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。

Wan 2.6 怎麼用
Wan 2.6 六種價格
Wan 2.6 如何選擇文字工作流程
Wan 2.6 單圖需要什麼規格
Wan 2.6 影片路由接收幾個檔案
Wan 2.6 哪些模式支援 15 秒
Wan 2.6 720p 1080p 點數
Wan 2.6 同時上傳圖片與影片會怎樣
Wan 2.6 提示詞最少幾個字元
Wan 2.6 鏡頭結構
Wan 2.6 內容過濾

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
先選定唯一輸入路徑

先選定唯一輸入路徑: 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。

2
依工作流程篩選時長

依工作流程篩選時長: 文字與圖片輸入支援 5、10 或 15 秒；影片輸入只支援 5 或 10 秒。

3
分別建立三條基線

分別建立三條基線: 預設為 1080p、5 秒、鏡頭結構關閉、內容過濾開啟。 請用相同提示詞、媒體、時長與解析度比較模型，並以實際結果驗證，不要預設行銷宣稱成立。

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 2.6 與 Kling 2.6 的差異
Wan 2.6 vs Kling 2.6 輸入路由與時長
Wan 2.6 與 Hailuo 2.3 的差異
Wan 2.6 vs Hailuo 2.3 素材與計費
Wan 2.6 與 Seedance 1.5 Pro 的差異
Wan 2.6 vs Seedance 1.5 Pro 控制項

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。

Kyeo 輸入路徑
Wan 2.6
Kyeo 輸入路徑 / Wan 2.6: 文字 / 1 張圖片 / 1–3 個影片
Wan 2.7 Video
文字、首尾幀或續寫、影片編輯分路由
Kling 2.6
文字或圖片兩條基礎路由
素材驅動三工作流程
Wan 2.6
素材驅動三工作流程 / 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。
Wan 2.7 Video
依素材切換 2.7 文字、圖片或影片編輯工作流程
Kling 2.6
依圖片是否存在切換文字或圖片路由
有效時長
Wan 2.6
有效時長 / Wan 2.6: 文字與圖片輸入支援 5、10 或 15 秒；影片輸入只支援 5 或 10 秒。
Wan 2.7 Video
文字／圖片支援 5/10/15 秒；影片編輯另有契約
Kling 2.6
文字與圖片均為 5/10 秒
六種 Kyeo 點數價格
Wan 2.6
六種 Kyeo 點數價格 / Wan 2.6: 720p 的 5/10/15 秒分別為 70/140/210 點，1080p 分別為 105/210/315 點。影片輸入不提供 15 秒。
Wan 2.7 Video
依 2.7 工作流程、解析度與時長另行計價
Seedance 1.5 Pro
依解析度、時長與輸入模式計價
可見控制與素材
Wan 2.6
可見控制與素材 / Wan 2.6: 解析度, 影片長度, 鏡頭結構, 內容過濾
Wan 2.7 Video
控制項與素材欄位依 2.7 工作流程變化
Kling 2.6
顯示 Kling 自身的時長、比例與聲音控制

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。

Wan 2.6 — 素材驅動三工作流程

Wan 2.6 / 素材驅動三工作流程: 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。

Wan 2.6 — 六種查表計費組合

Wan 2.6 / 六種查表計費組合: 720p 的 5/10/15 秒分別為 70/140/210 點，1080p 分別為 105/210/315 點。影片輸入不提供 15 秒。 請用相同提示詞、媒體、時長與解析度比較模型，並以實際結果驗證，不要預設行銷宣稱成立。

功能詳解

素材驅動三工作流程

素材驅動三工作流程: 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。

六種查表計費組合

六種查表計費組合: 720p 的 5/10/15 秒分別為 70/140/210 點，1080p 分別為 105/210/315 點。影片輸入不提供 15 秒。

素材與安全邊界

素材與安全邊界: 只接受 1 張至少 256×256 且不超過 10 MB 的 JPG/PNG/WebP 圖片，或 1–3 個每個不超過 10 MB 的 MP4/MOV 影片。開始任務前，每個檔案都必須通過格式、大小、尺寸與簽章中繼資料驗證。 鏡頭結構預設關閉，內容過濾預設開啟；兩個控制項都會顯示並明確送出。

適合做什麼

純文字基線

純文字基線: 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。

單圖動畫測試

單圖動畫測試: 只接受 1 張至少 256×256 且不超過 10 MB 的 JPG/PNG/WebP 圖片，或 1–3 個每個不超過 10 MB 的 MP4/MOV 影片。開始任務前，每個檔案都必須通過格式、大小、尺寸與簽章中繼資料驗證。 預設為 1080p、5 秒、鏡頭結構關閉、內容過濾開啟。

1–3 MP4/MOV — 影片長度

1–3 MP4/MOV — 影片長度: 只接受 1 張至少 256×256 且不超過 10 MB 的 JPG/PNG/WebP 圖片，或 1–3 個每個不超過 10 MB 的 MP4/MOV 影片。開始任務前，每個檔案都必須通過格式、大小、尺寸與簽章中繼資料驗證。 文字與圖片輸入支援 5、10 或 15 秒；影片輸入只支援 5 或 10 秒。

提示詞技巧

先確認唯一素材路徑

先確認唯一素材路徑: 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。

每次只更改一個輸出參數

每次只更改一個輸出參數: 720p 的 5/10/15 秒分別為 70/140/210 點，1080p 分別為 105/210/315 點。影片輸入不提供 15 秒。

鏡頭結構 / 內容過濾

鏡頭結構 / 內容過濾: 鏡頭結構預設關閉，內容過濾預設開啟；兩個控制項都會顯示並明確送出。 模型不會生成音軌。每次都要檢查結果的安全性、權利、身分一致性與發布品質。

選擇它的理由

功能標籤: 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。
點數消耗: 720p 的 5/10/15 秒分別為 70/140/210 點，1080p 分別為 105/210/315 點。影片輸入不提供 15 秒。
上傳限制: 只接受 1 張至少 256×256 且不超過 10 MB 的 JPG/PNG/WebP 圖片，或 1–3 個每個不超過 10 MB 的 MP4/MOV 影片。開始任務前，每個檔案都必須通過格式、大小、尺寸與簽章中繼資料驗證。
影片長度: 文字與圖片輸入支援 5、10 或 15 秒；影片輸入只支援 5 或 10 秒。

使用前須知

Wan 2.6 — 依工作流程篩選時長: 文字與圖片輸入支援 5、10 或 15 秒；影片輸入只支援 5 或 10 秒。
Wan 2.6 — 上傳限制: 只接受 1 張至少 256×256 且不超過 10 MB 的 JPG/PNG/WebP 圖片，或 1–3 個每個不超過 10 MB 的 MP4/MOV 影片。開始任務前，每個檔案都必須通過格式、大小、尺寸與簽章中繼資料驗證。
Wan 2.6 — 先選定唯一輸入路徑: 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。
Wan 2.6 — 鏡頭結構 / 內容過濾: 鏡頭結構預設關閉，內容過濾預設開啟；兩個控制項都會顯示並明確送出。 模型不會生成音軌。每次都要檢查結果的安全性、權利、身分一致性與發布品質。

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：1080p
720p
1080p
影片長度
選填
參數類型: select
預設值：5
5 秒
10 秒
15 秒（僅文字／圖片）
鏡頭結構
選填
參數類型: 選擇
預設：false
單鏡頭（false）
多鏡頭（true，最多 5 個）
內容過濾
選用
參數類型: 下拉選擇
預設：開啟
關閉
開啟

點數消耗

70–315 點數／次

720p 的 5/10/15 秒分別為 70/140/210 點，1080p 分別為 105/210/315 點。影片輸入不提供 15 秒。

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。

Kling 2.6

Kling 2.6 在 Kyeo 是一項雙工作流程影片服務：沒有圖片時呼叫文字生成影片，有圖片時呼叫單圖圖片生成影片。兩路都提供 5/10 秒與 `sound` 開關；文字工作流程另有三種長寬比。四個計費組合為 55、110、110、220 點數。

AI 影片模型
每次 55-220 點數
查看詳情立即使用

Hailuo 2.3

Kyeo 有六種有效組合：Standard 的 768P 6/10 秒與 1080P 6 秒為 25/40/40 點數；Pro 對應為 40/80/70 點數。1080P 不支援 10 秒。 必填 1 個 JPG/PNG/WebP 檔案，最大 10MB.

AI 影片模型
每次 25-80 點數
查看詳情立即使用

Seedance 1.5 Pro

Seedance 1.5 Pro / 功能標籤 — Kyeo 的 Seedance 1.5 Pro 支援純文字或最多 2 張圖片輸入，提供 480p／720p／1080p、4–12 秒、生成音訊、6 種比例與鎖定鏡頭。7–180 點數由參數組合決定；Pro 名稱與較高點數不代表本頁已驗證畫質、穩定性或發布可用性。

AI 影片模型
7–180 點數／次
查看詳情立即使用

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

本頁依據 Alibaba Cloud 對 Wan 2.6 系列的官方發布資訊與官方影片模型概覽完成複核。以下公開事實會區分文字、圖片與影片輸入、可用時長、站內點數、已驗證的媒體限制和可見控制項；行銷宣稱不視為實測結果。

最後更新: 2026-08-27
Alibaba Cloud: Wan 2.6 series launch
官方
查看原始來源
Alibaba Cloud: video generation model overview
官方
查看原始來源