Text and frame control
Create from text alone, animate a first frame, or add an ending frame in Canvas. Frame control and general references are separate input workflows.
Wan 3.0
Turn a written scene, an opening frame or a set of references into video with Wan 3.0. Direct the action, camera and sound in one creative workflow.
Wan 3.0
About Wan 3.0
Wan 3.0 is Alibaba’s video generation model for scenes that need time to unfold. Start from a written idea, animate an opening image, or bring visual and audio references into a new scene. A continuous camera move, a character’s changing expression or a carefully timed reveal can develop within one generation of up to 30 seconds.
Reference material gives your direction a visual starting point. Use images to establish a character, object or setting, and video to guide movement or camera rhythm. For an approved opening and ending composition, use first- and last-frame control in Canvas. Describe what should happen between the frames rather than repeating everything already visible.
Arcloop connects the standard Wan 3.0 model. Choose 2–30 seconds, 480p / 720p / 1080p and a landscape, portrait or square format in Canvas. Generated audio is optional. Start with the composer on this page, then continue in Canvas for reference roles, output settings and the current credit estimate.
Wan 3.0 Video Prompt Examples
Generated with Wan 3.0 at 720p for 5–6 seconds. View the exact prompt used for each video.
5秒，写实军事悬疑电影片段，三个镜头硬切，明确分镜，不使用渐变转场。
同一名35岁左右的欧美白人成年男性狙击手，棕色短发、浅色眼睛、自然的短胡茬、轮廓分明的面部，穿灰绿色战术外套和低调的战术背心。他在一座空旷旧城区的屋顶上，利用低矮的混凝土护墙隐蔽，手持装有瞄准镜的长枪。自然、真实的皮肤质感，安静而紧张，人物外貌和服装跨镜头一致。
镜头1，0–1.5秒：24mm广角中远景，从屋顶侧后方拍摄。男子半跪在护墙后，面向远处的城市。傍晚阴天，冷色天空映出屋顶轮廓，风吹动衣角。画面交代他的身体位置、护墙和远方街道。结束时直接硬切。
镜头2，1.5–3.5秒：侧面中近景，平视机位。清楚展示男子的欧美面部轮廓、短胡茬、肩部和瞄准镜。他将脸颊轻轻贴近枪托，身体保持稳定，目光专注。浅景深，远处楼群柔和虚化。保持与上一镜头一致的观察方向。结束时直接硬切。
镜头3，3.5–5秒：从斜前方切入紧凑的眼神特写，清晰看见同一名男子专注的浅色眼睛、微皱眉头和瞄准镜边缘。他轻缓呼气，目光保持不动。用克制的表情和皮肤细节结束画面。
真实电影摄影，低饱和冷灰色调，柔和自然侧光，暗部有细节，轻微胶片颗粒。每个镜头构图稳定，切镜清楚，人物与枪械形态一致。人物始终是成年人，欧美白人面孔，不使用真实演员或公众人物的相貌。
声音：屋顶风声、远处城市低鸣、衣料摩擦、最后一次轻缓呼气。无对白、无背景音乐、无字幕。镜头只表现观察和准备，没有射击、命中画面或血腥内容。
5秒，单个连续镜头，手工黏土定格动画风格。
一名圆头、短腿的薄荷绿色黏土小人，穿着浅黄色背带裤，站在迷你菜园里，双手紧紧抓着一颗胡萝卜的叶子。人物有简单的黑色圆眼睛和小嘴，黏土表面保留细小指纹、手工压痕和柔软的边缘。土壤、植物和人物都由真实感黏土制成。
0–2秒：小人身体向后倾，用力拉住萝卜叶，双脚牢牢踩在地上，脸颊因使劲微微鼓起。
2–4秒：胡萝卜突然从土中拔出，带起几粒黏土土块。小人失去平衡，一屁股坐在地上，仍然抱着萝卜。
4–5秒：小人愣一下，低头看看怀里的大萝卜，露出满足的笑容。
固定的低机位中景，人物全身和萝卜始终留在画面内。温暖柔和的摄影棚光线，浅景深，精致的微缩布景。动作带有手工定格动画的节奏，人物造型和材质始终一致。
声音：轻微用力的哼声、萝卜拔出的“啵”声、柔软的落地声。无对白、无背景音乐、无字幕和标识。
写实电影风格的美国西部片，19世纪末，一座荒凉的边境小镇，暴风雨前的深灰色天空，低角度夕阳穿过风沙，照亮木制酒馆和尘土飞扬的街道。保持同一名主角的面容、服装和位置关系连续：一位四十岁左右、欧美面孔、胡茬明显的赏金猎人，戴磨旧的深棕色牛仔帽，穿褪色长风衣，腰间佩戴左轮手枪。
0–2.5秒，低机位中景：镜头从街道另一端缓慢向赏金猎人推进。他独自站在街中央，侧身面对酒馆，右手自然垂在枪套旁。风卷起脚边的尘土，风衣下摆摆动，一团风滚草从前景横穿画面。
2.5–4秒，硬切至侧面近景：赏金猎人缓慢抬起下巴，露出帽檐下警惕的双眼。他的视线始终朝向画面右侧，嘴角微微收紧。背景虚化的酒馆双扇门被人推开。
4–6秒，硬切反打至酒馆门口：从赏金猎人肩后拍摄，一名穿黑色马甲的陌生人停在门框阴影里，两人隔着空旷街道对视。镜头保持稳定，以悬而未决的对峙结束。三个镜头保持同一场景、同一时刻和清晰的空间方向。
变形宽银幕电影镜头质感，克制的胶片颗粒，真实皮肤、皮革与木材纹理，暖色夕阳与冷色乌云形成层次。原生同步音频：持续的低沉风声和砂砾滚动声，风衣轻微扑动声，皮靴踩过砂砾的轻响，酒馆木门打开时短促的吱呀声，结尾传来一声远处低沉的雷鸣。无对白、无配乐、无字幕。全片不拔枪、不射击，用眼神、空间和声音建立紧张感。
5秒，写实电影风格的公路片，三个清晰的硬切分镜。初秋傍晚的加州沿海公路，金色斜阳，远处是蓝色海面、海岸峭壁和低矮的干草山坡。一辆复古焦橙色双门轿跑车，米色皮革内饰，驾驶侧车窗打开，沿着右侧车道平稳行驶。全片始终是同一辆车、同一名驾驶者和同一段旅程。驾驶者是一位三十岁左右、欧美面孔、棕红色齐肩头发的成年女性，穿米白色衬衫，系好安全带，独自开车，目光始终关注前方道路。
0–1.8秒，车外前侧方跟拍中远景：摄影机与汽车同速平行移动，车辆从画面左侧向右侧前进。焦橙色车身上掠过温暖的夕阳反光，车轮自然旋转，海岸和路旁草坡在背景中连续滑过。车辆始终贴合路面，运动流畅、速度适中。
1.8–3.4秒，硬切至车内副驾驶位置的驾驶者侧面近景：镜头稳定，清楚看到她的侧脸、前方挡风玻璃和方向盘。她双手自然扶住方向盘，轻轻调整方向，目视前方，嘴角露出放松的微笑。打开的侧窗吹动几缕头发，夕阳在她的面颊和仪表台上移动。背景道路的运动方向与上一镜头一致。
3.4–5秒，硬切至车后方跟随远景：同一辆焦橙色轿跑车沿海岸公路驶入一个宽缓弯道，摄影机逐渐落后，让车辆在海岸和开阔海面之间向远处驶去。以自由、安静而略带怀旧的旅途氛围结束。
35毫米电影胶片质感，自然的皮肤和车漆纹理，柔和的高光，克制的镜头光晕，真实的车轮运动模糊。保持人物面容、服装、车辆造型和光照连续，驾驶动作清晰可信。原生同步音频：低沉平稳的汽油发动机声、轮胎与沥青的摩擦声、侧窗风声，切入车内时声音自然变得低闷，切回车外时恢复开阔的风声。无对白、无配乐、无字幕或标识。
电影级日式二维动画，雨后黄昏的小镇车站，精细手绘背景，蓝紫色暮光与车厢暖光。黑色齐肩短发少女，深蓝校服外套，米白帆布包。
0–1.8秒：侧面跟拍少女沿站台小跑。
1.8–3.3秒：硬切近景，她停下喘气，抬头望向右侧，神情从焦急变为惊讶，晚风吹动碎发。
3.3–5秒：硬切少女肩后反打，黑发少年站在敞开的列车门内，回头与她对视，轻轻微笑。
列车始终静止，人物造型和视线方向连续，动作自然，情绪克制。声音：脚步、轻微喘息、晚风和列车低鸣。无对白、配乐、字幕。
Wan 3.0 capabilities
Multimodal references
Canvas supports up to 10 images, 5 videos and 5 audio clips. Reference video and audio each allow 15 seconds in total; reference video plus output must fit within 30 seconds.
Quality–cost balance
Balance visual quality, motion, and native sound with an accessible generation cost.
How to use Wan 3.0
Start here and continue in Arcloop Canvas to choose the final inputs, settings and credit estimate.
- 01
Describe the scene
Write the subject, action and camera direction. Add an optional image or video here; organize additional references in Canvas.
- 02
Continue to Canvas
Click Generate Now and sign in if needed. Your prompt and selected reference continue into a Story Canvas.
- 03
Choose the inputs and settings
Set the reference roles, aspect ratio, resolution, duration and audio. Check the current credit estimate before submitting.
- 04
Generate the video
Submit from Canvas when the setup is ready. The result stays with your story for the next part of the production.
What you can create
Short dramas
Character encounters, compact dialogue scenes and emotional turning points.
Film scenes
Continuous camera moves, action beats and atmospheric trailer moments.
Product films
Controlled product reveals and brand stories, from realism to stylized animation.
Tips for Wan 3.0
Practical guidance adapted from Alibaba’s Wan 3.0 prompt guide and the controls available in Arcloop.
- 01
Write the action in order
For one shot, name the main action and the camera move. For several beats, give each a time range that fits the selected duration.
- 02
Give references a purpose
Say which image defines the character or location and which video guides motion. Keep first/last-frame control separate from general references.
- 03
Specify dialogue and sound
Name the speaker, put spoken words in quotation marks, then describe ambience, effects and music. Keep dialogue short enough for the clip.
- 04
Keep constraints selective
Add only the exclusions you need, such as no subtitles or no background music. Put them in the prompt and avoid repeating or conflicting instructions.
Wan 3.0 Pricing
Wan 3.0
Generation settings: Variable settings
Reference credits: Based on actual usage
Prices are estimates. Check the Canvas quote for your selected resolution, duration and references; actual charges follow final settlement.
Wan 3.0 alternatives
Choose an alternative by the inputs and shot you want to create. Settings below describe the current Arcloop integrations.
Wan 3.0
Balance visual quality, motion and native sound with generation cost, using text, frame control or multimodal references.
- Duration
- 2–30 seconds
- Reference limits
- 10 images / 5 videos / 5 audio
- Output resolution
- 480p / 720p / 1080p
Kling 4.0
Reference-led character performance, deliberate camera movement and short-drama or action scenes.
- Duration
- 3–30 seconds
- Reference limits
- 10 images / 5 videos / 3 audio; 15 total
- Output resolution
- 720p / 1080p
Seedance 2.5
Larger reference sets, connected story beats and films assembled from multiple shots.
- Duration
- 4–30 seconds
- Reference limits
- 30 images / 10 videos / 10 audio
- Output resolution
- 480p / 720p / 1080p
FAQ
What can I use as input?
Start from text, use a first frame with an optional last frame, or combine general image, video and audio references in Canvas. First/last-frame control cannot be mixed with general references in the same request.
How many references can I use?
Canvas supports up to 10 images, 5 videos and 5 audio clips. Reference videos total up to 15 seconds and reference audio totals up to 15 seconds. Reference-video duration plus output duration cannot exceed 30 seconds. The quick composer on this page is a starting input; use Canvas to organize the full set.
Can Wan 3.0 generate sound?
Yes. Generated audio is optional and starts enabled here. Describe the dialogue, environmental sound or music you want, then confirm the audio setting in Canvas.
How are credits calculated?
Resolution, output duration and reference-video usage affect the quote. Canvas shows the available settings and credit estimate before submission; actual charges follow final settlement.
Start creating with Wan 3.0
Turn a written scene, an opening frame or a set of references into video with Wan 3.0. Direct the action, camera and sound in one creative workflow.