What Is Wan? Alibaba's Video & Image Stack
2026-11-10
TLDR
Wan is Alibaba's family of video and image models, not a single tool for every visual job. AIAdapter currently lists three Wan video models: Wan 3.0 Video, Wan 3.0 Video Prime, and Wan 2.7 VideoEdit. Wan image models are a separate branch; choose a listed image model when you need a still, or a Wan video model when you need motion.
Want to turn a product photo into a clip, or make a new scene from a few lines of text? Wan is Alibaba's family of AI models for video and images. Start with the asset you need: the models and inputs for a moving clip differ from those for a still image. Once you know the output, you can choose the model and API operation that fit the job.
Alibaba's Wan lineup spans text-to-video, image-to-video, reference-guided video, video editing, text-to-image, and image editing. AIAdapter lists Wan video models in its model marketplace; the family's image models are a separate branch.
Wan video and Wan image solve different jobs
A video model turns a prompt or visual input into moving footage. Start from text when you need to invent a scene. Start from an image when you already have a product shot, illustration, or character frame whose appearance should carry into a clip. Reference-guided generation brings additional source material into the creative direction; video editing starts from footage you want to change.
An image model produces a still. Alibaba Cloud describes wan2.6-t2i for text-to-image and wan2.6-image for image editing and interleaved text-image output. These upstream names are not AIAdapter Wan video IDs. An image-to-video model accepts an image but returns a clip, not an edited photograph. For a product campaign, create or select the still first, then animate it; editing the still itself requires an image operation.
Which Wan models can you explore on AIAdapter?
AIAdapter's marketplace currently presents three Wan entries under the aliyun-bailian vendor prefix: Wan 3.0 Video, Wan 3.0 Video Prime, and Wan 2.7 VideoEdit. The full prefix is part of the model ID you send in an AIAdapter request.
| Your output | Model to explore | Why it fits the task |
|---|---|---|
| A new short clip from a written concept | aliyun-bailian/wan3.0-video | The listed Wan 3.0 video option; its model page shows a prompt-based video-generation request. |
| Another Wan 3.0 video option for the same brief | aliyun-bailian/wan3.0-video-prime | A separately listed video model that you can compare using the same creative brief and current rate information. |
| Changes to existing video footage | aliyun-bailian/wan2.7-videoedit | The listed Wan video-editing variant; editing requires source footage and a model-specific input format. |
| A new or edited still image | An image model in the marketplace | Use an image-generation or image-editing model rather than a Wan video endpoint. |
Alibaba Cloud documents both Wan 3.0 video variants as all-in-one models for several video tasks, including text-to-video and video guided by frames or references. On AIAdapter, the model pages show a text-prompt request, a straightforward way to make a first clip. If you have frames, reference media, or footage to work from, prepare them for the API operation you select.
Alibaba Cloud positions the upstream Wan 3.0 Video Prime variant as a high-speed version. That description does not establish how fast a finished request will be on AIAdapter, nor does it guarantee better image quality. Try the same scene brief with the two listed IDs, then compare usable clips, completion times, and the current billed rates for your settings.
Where Wan fits a creative workflow
Product storytelling from an existing visual
Suppose you have a product photo with a color and silhouette the customer should recognize. Use it as the visual starting point for a tabletop shot or a slow camera move around the product. Tell the model which details should stay recognizable, what should move, and where the shot should end. For ecommerce, compare the candidate clips with the brand-approved product photo and choose the one that shows the product most faithfully.
New scenes for ads and social video
When no starting image exists, text-to-video lets you test a visual idea. Specify subject, action, shot size, camera movement, lighting, and the ending. Use the same brief to compare which outputs communicate the message. Alibaba also describes multi-shot storytelling and audio-video synchronization in suitable Wan variants; add reference or sound requirements when the selected model and request mode support them.
Revision rather than regeneration
Already have a clip you like except for one object in the frame? Wan 2.7 VideoEdit is the listed option for revising existing footage. Give it the source footage and describe the change, such as replacing that object while keeping the action. Compare the original and edited versions to see whether the details you wanted to retain are still there.
How a Wan video API call works on AIAdapter
AIAdapter documents a common API base at https://api.aiadapter.com/v1 and a video workflow with two steps: submit a generation job through POST /v1/video/generations, then check it through GET /v1/video/generations/{id}. Use a server-side API key and the exact marketplace model ID. For a first text-to-video test, the Wan 3.0 Video model page shows this request pattern:
curl https://api.aiadapter.com/v1/video/generations \
-H "Authorization: Bearer $AIADAPTER_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"aliyun-bailian/wan3.0-video","prompt":"A close-up of a ceramic cup on a sunlit table as the camera slowly pulls back."}'Use the job ID returned by the submission to check progress:
curl https://api.aiadapter.com/v1/video/generations/JOB_ID \
-H "Authorization: Bearer $AIADAPTER_API_KEY"Video generation is asynchronous: the first response identifies a job, not necessarily a finished clip. The video API documentation covers submission and polling. This text-only request does not send a source image or video; image-to-video, reference-guided generation, and editing may need different media fields. Follow the AIAdapter operation for your selected model rather than copying an Alibaba Cloud direct-API payload unchanged.
With AIAdapter's common API base and the full model ID, your team can try the listed Wan video options from the same integration. For a photo-led clip or an edit, supply the image or source footage in the format required by that task.
How to choose a model without comparing names alone
Start with the asset: still, new clip, animated still, or edited footage. For video, use a repeatable brief and judge whether the subject stays recognizable, the motion communicates the idea, and the result needs repair. Cost per usable clip matters more than a headline rate when a shot takes several attempts.
AIAdapter's Wan video listings show per-second rates. Its pricing page explains the 15% service fee on the discounted cost. Check the selected model's current billed rate and your requested settings; still images follow a different billing unit.
If you need still-image creation now, the marketplace also lists Qwen-Image-2.0-Pro as an image model. It belongs to Alibaba's Qwen-Image family, not the Wan family. That is a useful distinction when planning an end-to-end asset workflow on AIAdapter: choose a listed image model for the still, and a listed Wan video model for motion.
Frequently asked questions
Is Wan one model or a family of models?
Wan is a family. Alibaba offers distinct models and operations for generating video from text, creating video from images or references, editing footage, and making or editing still images. A family name tells you the creative domain, but not which request to send. For an integration, identify the output you need first, then use the exact model ID and API operation offered by your chosen service. On AIAdapter, the currently listed Wan entries are video-related, so an image task calls for a separate image model.
Is Wan the same as Qwen-Image?
No. Both are associated with Alibaba's broader AI portfolio, but they name different model families. Wan includes video and image-generation branches. Qwen-Image is a distinct family focused on image creation and editing. A team can use a Qwen-Image model to create a still and a Wan video model to animate or build a clip from its visual direction, but the image-model ID cannot replace a Wan video-model ID in a video request.
Can Wan turn a product photo into a video?
Alibaba documents image-to-video among Wan's video tasks. The starting photo serves as a frame or visual guide, and the intended output is footage. For product work, make the prompt specify which details must remain recognizable and what motion should happen. The exact way to attach that photo depends on the selected model's API. A prompt-only video submission illustrates the workflow but does not itself send an image; use the image-input fields supported by the service you actually call.
What is the difference between Wan 3.0 Video and Wan 3.0 Video Prime?
AIAdapter lists them under separate IDs: aliyun-bailian/wan3.0-video and aliyun-bailian/wan3.0-video-prime. Alibaba Cloud calls the upstream Prime variant a high-speed version, but that label alone cannot tell you the end-to-end completion time on AIAdapter or which result suits your scene. Send the same brief to both listed models and compare usable footage, completion time, and the billed rate for your settings.
Does an OpenAI-compatible API mean the same request works for every Wan task?
No. The common API base lets you access listed models from one integration, but each task still calls for the right source material: a prompt for text-to-video, an image for image-to-video, or source footage and an instruction for video editing. Still-image generation calls for an image operation, not the video job endpoint. Choose the model and API operation together, then provide the material that task needs.
Ready to try a Wan video concept? Create an AIAdapter account, pick a listed Wan model, and begin with one short, clearly defined scene.
aliyun-bailian/wan3.0-video-prime
aliyun-bailian/wan2.7-videoedit
FAQ
What Is Wan? Alibaba's Video & Image Stack
Wan is Alibaba's family of video and image models, not a single tool for every visual job. AIAdapter currently lists three Wan video models: Wan 3.0 Video, Wan 3.0 Video Prime, and Wan 2.7 VideoEdit. Wan image models are a separate branch; choose a listed image model when you need a still, or a Wan video model when you need motion.