Alibaba
Wan API
Alibaba's Wan line — one family covering both image and video generation.
3
models in the line
Wan (Wanxiang) is Alibaba's generation line, and unlike the other families here it spans both media types: one image model and two video models. The video side is built around reference work — first frame, first and last frame, multimodal references and continuation of an existing clip — through a single flexible model rather than a separate endpoint per mode.
At a glance
- From
- 0.029 coins / generation
- Modes
- Text to image, Image to image, Text to video, Image to video
- Resolutions
- 480p, 720p, 1080p
- Aspect ratios
- 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, adaptive
Models in this family
| Model | From | Modes | Open |
|---|---|---|---|
| Wan 2.7 ImageGenerate and edit images with aspect-ratio, pixel-size, and seed controls. | 0.029 coins/ generation | Text to image, Image to image | OpenDocs |
| Wan 3.0 VideoAlibaba Wanxiang 3.0 all-in-one reference video model. Text, first frame, first + last frame, multi-modal reference (images, video, audio) and document or web page references, 2-30 seconds at 480p/720p/1080p. | 0.05 coins/ second | Text to video, Image to video | OpenDocs |
| Wan 2.7 VideoGenerate, edit, reference, or continue video through one flexible model. | 0.09 coins/ second | Text to video, Image to video | OpenDocs |
Pricing
| Model | Option | Price |
|---|---|---|
| Wan 2.7 Image | Standard | 0.029 coins/ generation |
| Wan 3.0 Video | 480p | 0.05 coins/ second |
| Wan 3.0 Video | 720p | 0.099 coins/ second |
| Wan 3.0 Video | 1080p | 0.198 coins/ second |
| Wan 2.7 Video | 720p | 0.09 coins/ second |
| Wan 2.7 Video | 1080p | 0.135 coins/ second |
Which one to pick
Wan 2.7 Image
Image generation and editing with aspect-ratio, pixel-size and seed control — the seed makes runs reproducible.
Wan 3.0 Video
The cheaper and newer video model, with an explicit frame-versus-reference mode switch and 480p through 1080p.
Wan 2.7 Video
Video work that mixes generation, editing, reference and continuation in one model, at 720p or 1080p.
Strengths
- One family for image and video, so a single vendor relationship covers both.
- Seed control on the image model, which makes a generation reproducible.
- Wan 3.0 exposes the reference mode explicitly rather than inferring it from which fields you sent.
- Video starts at 480p, which keeps draft renders cheap.
Limitations
- The family is small — one image model and two video models, with no cheap bulk tier on the image side.
- Wan 2.7 video starts at 720p, so the cheapest video drafts have to come from 3.0.
- The two video models differ in parameters, so they are not interchangeable in code.
The vendor's own description of this line: Alibaba Cloud Model Studio — video generation
Guides
- Image generationWhich image model to use for the picture you need, prompts that produce it, batching for variants, and the API run from prompt to file URL.8 min read
- Image editingEditing photos with references instead of prompts — background swaps, product composites, style matching, and the Midjourney action chain from imagine to inpaint.8 min read
- GPT Image 2.5 — Sunburst, Flare, prompts and real examplesChoose a GPT Image model, compare Sunburst and Flare, copy four practical prompts, inspect six real renders, and generate or edit images through API Hubs.9 min read