| --- |
| license: apache-2.0 |
| base_model: Wan-AI/Wan2.2-Animate-14B |
| language: |
| - en |
| tags: |
| - video-to-video |
| - image-to-video |
| - character-animation |
| - motion-transfer |
| - wan |
| - wan2.2-animate |
| - api |
| - apimodels |
| --- |
| |
| # Wan2.2-Animate — hosted API access |
|
|
| This repository is **not the model weights** — it documents running **Wan2.2-Animate** through a |
| hosted API on **[apimodels.app](https://apimodels.app)**, a multi-model gateway where one key and one |
| OpenAI-compatible endpoint reach 140+ image, video, audio and language models. |
|
|
| - Original model & weights: **https://huggingface.co/Wan-AI/Wan2.2-Animate-14B** |
| - Authors: the Wan team (Alibaba Tongyi) — paper: *Wan: Open and Advanced Large-Scale Video Generative Models* (arXiv:2503.20314) |
| - License: Apache-2.0 |
|
|
| Independent hosted-API integration. Not affiliated with or endorsed by the authors; does not |
| redistribute their weights. All credit for the model is theirs. |
|
|
| ## What it does |
|
|
| Give it a **character image** and a **reference video**, and it returns a new clip in one of two modes: |
|
|
| - **animate** (default): your character and background are driven by the motion and expressions of |
| the person in the video — dance covers, gestures, a talking head — no motion capture needed. |
| - **replace**: the video keeps its scene, camera and motion, and only the person is swapped for your |
| character. |
|
|
| The output is exactly as long as the reference video (up to 120 seconds), at 480p or 720p, 24 fps, |
| and it keeps the reference video's audio track. |
|
|
|  |
|
|
| *Left: the character image (from the official Wan2.2 example assets). Right: a frame of the clip produced |
| through the apimodels.app API in animate mode — the character follows the gestures of the reference video.* |
|
|
| ## Use it via API |
|
|
| On apimodels.app the model is served as **wan-2.2-animate**. Get a key at https://apimodels.app, then: |
|
|
| ```bash |
| # image + reference video -> video (async): create, then poll |
| curl -X POST https://api.apimodels.app/v1/video/generations \ |
| -H "Authorization: Bearer YOUR_API_KEY" -H "Content-Type: application/json" \ |
| -d '{ "model": "wan-2.2-animate", |
| "image": "https://example.com/character.png", |
| "video": "https://example.com/dance-template.mp4", |
| "mode": "animate", |
| "resolution": "480p" }' |
| # -> { "data": { "taskId": "..." } } |
| curl "https://api.apimodels.app/v1/video/generations?task_id=TASK_ID" \ |
| -H "Authorization: Bearer YOUR_API_KEY" # state -> completed, resultUrls[] |
| ``` |
|
|
| Parameters: `image` (public URL or base64, up to 50 MB), `video` (public https MP4 URL, up to 120 s |
| and 200 MB), `mode` (`animate` | `replace`), `resolution` (`480p` | `720p`, default 720p), optional |
| `prompt` and `seed`. There is no duration parameter — the output follows the reference video. |
|
|
| ## Pricing |
|
|
| Billed per second of the **reference video**, rounded up to whole seconds, fixed when the task is created: |
|
|
| | Tier | Price | |
| | --- | --- | |
| | 480p | $0.0742 / second | |
| | 720p | $0.154 / second | |
|
|
| A 5-second reference costs $0.371 (480p) or $0.77 (720p). Both modes cost the same. Failed jobs are |
| not charged. Result files are kept for 7 days. |
|
|
| ## Links |
|
|
| - Run it: **[apimodels.app](https://apimodels.app/models/wan-2.2-animate)** |
| - API docs: https://apimodels.app/docs/wan-2-2-animate |
| - Original model: https://huggingface.co/Wan-AI/Wan2.2-Animate-14B |
|
|
