Unlimited Open Source Models

Get Plan
Skip to main content
Available now on ModelsLab · Video Generation

Wan 2.2 API — Text-to-Video & Image-to-VideoWan 2.2, the successor to Wan 2.1, on ModelsLab GPUs. $0.075 per clip, unlimited on the $149 plan.

Try Wan 2.2Wan 2.7 APIAI Video API

Wan 2.2 API: endpoints, output, speed and price

Price first

$0.075 per clip

Wan 2.2 runs on ModelsLab's own GPUs at $0.075 per generation, drawn from your plan's included usage ($21 on Basic, $47 on Standard). The $149 Open Source Unlimited plan has no limit on it.

Text-to-Video

wan-2.2-t2v

POST a prompt to /api/v6/video/text2video_ultra with model_id wan-2.2-t2v. Leave portrait off for a 16:9 clip or turn it on for 9:16.

Image-to-Video

wan-2.2-i2v

POST an init_image and a motion prompt to /api/v6/video/img2video_ultra with model_id wan-2.2-i2v. The subject and framing of the still carry into the clip.

Resolution and length

480p, about 5 seconds

The hosted endpoints render 480p MP4 or MOV. num_frames 82-128 at 16-20 fps gives about 4 to 8 seconds; 92 frames at 18 fps is about 5 seconds.

Generation time

About 45-60 s median

Median wall-clock time on ModelsLab over the last 30 days: about 46 seconds for text-to-video and 61 seconds for image-to-video, queue included. The slowest 10% take 4-5 minutes, so use a webhook or the fetch endpoint.

Wan 2.2 vs Wan 2.1

Drop-in successor

Wan 2.2 keeps the Wan 2.1 architecture and prompts with better motion coherence. model_id wan2.1 is still available on the same text-to-video endpoint.

Wan 2.7 API

Need image edit too?

Wan 2.7 adds text-to-image and image edit to the same family and runs on Alibaba Cloud, so it is billed per call as a third-party model and is not covered by Open Source Unlimited. See the Wan 2.7 API page.

Webhooks

Async callbacks

Pass a webhook URL — ModelsLab POSTs the MP4 to your endpoint when generation completes. Or poll POST /api/v6/video/fetch/{id}.

Endpoints

Multiple endpoints, one model

Wan 2.2 API — Text-to-Video & Image-to-Video is available through 2 specialized endpoints. Choose the one that fits your workflow.

Wan Text to Video Ultratext-to-video-ultraWan Image to Video Ultraimage-to-video-ultra

Examples

See what Wan 2.2 can do

Copy any prompt below and try it yourself in the playground.

Cinematic landscape

“a cinematic shot of a red fox walking through snow, depth of field, golden hour light”

Product reveal

“a sleek smartphone rotating on a minimalist pedestal, studio lighting”

Character animation

“a dancer performing fluid contemporary moves in a sunlit studio”

Nature scene

“a waterfall crashing into a pool at dusk, natural ambient motion”

For Developers

A few lines of code.
One prompt, $0.075, about 5 seconds of video.

ModelsLab handles the infrastructure: fast inference, auto-scaling, and a developer-friendly API. No GPU management needed.

  • Serverless: scales to zero, scales to millions
  • Pay per generation, no minimums
  • Python and JavaScript SDKs, plus REST API
API Documentation 
import requests


response = requests.post(
    "https://modelslab.com/api/v6/video/text2video_ultra",
    json={
  "fps": "18",
  "key": "YOUR_API_KEY",
  "prompt": "a red fox walking through fresh snow at golden hour, shallow depth of field, slow tracking shot",
  "model_id": "wan-2.2-t2v",
  "portrait": false,
  "num_frames": "92",
  "resolution": "480",
  "output_type": "mp4",
  "negative_prompt": "blurry, low quality, distorted, watermark, text"
}
)
print(response.json())

FAQ

Common questions about Wan 2.2 API — Text-to-Video & Image-to-Video

Read the docs

The Wan 2.2 API runs Alibaba's open-weight Wan 2.2 video model on ModelsLab's own GPUs. A generation costs $0.075 from your plan's included usage (plans from $21/month), and the $149/month Open Source Unlimited plan has no limit on it. POST a prompt to /api/v6/video/text2video_ultra (model_id wan-2.2-t2v) or an image to /api/v6/video/img2video_ultra (model_id wan-2.2-i2v).

$0.075 per generation (model page, checked 2026-09-23), drawn from the usage included in your plan: $21 on Basic covers about 280 clips and $47 on Standard about 626. Open Source Unlimited ($149/month) has no limit on self-hosted open-weight models such as Wan 2.2. Kling, Veo and Seedance are third-party: they are paid from the same included usage on Basic and Standard, then from wallet balance, and are not covered by Open Source Unlimited.

Create a ModelsLab account, subscribe to a plan (from $21/month), and copy your API key from the dashboard. Send it as "key" in the JSON body. The same key works for every other image, video, audio and LLM model.

Text-to-video (wan-2.2-t2v on /api/v6/video/text2video_ultra) builds the whole clip from a prompt; set portrait for 9:16. Image-to-video (wan-2.2-i2v on /api/v6/video/img2video_ultra) animates an init_image you supply, so the subject and framing stay fixed and the prompt only describes the motion.

The hosted endpoints render 480p. Length is num_frames divided by fps: 82-128 frames at 16-20 fps, which is about 4 to 8 seconds. 92 frames at 18 fps gives about 5 seconds. For longer videos, generate clips and join them at scene boundaries.

Over the last 30 days on ModelsLab, the median wall-clock time was about 46 seconds for text-to-video and 61 seconds for image-to-video, including queue time. The slowest 10% took 4-5 minutes, so pass a webhook or poll POST /api/v6/video/fetch/{id} instead of holding the request open.

Wan 2.2 is Alibaba Cloud's upgrade over Wan 2.1 — same motion model lineage, backward-compatible prompt structure, better temporal coherence on longer clips. Most prompts that worked on 2.1 produce equal-or-better output on 2.2 without code changes.

Yes. Pass model_id wan2.1 to /api/v6/video/text2video_ultra. Use Wan 2.2 unless you need to reproduce output from existing Wan 2.1 prompts and settings.

You can run the open weights yourself in Colab or on your own GPU, but you then manage VRAM, model downloads, queues and uptime. The API is one HTTP call per clip at $0.075, or unlimited on the $149 plan, with no download and no GPU to keep warm.

Use Wan 2.2 for bulk text-to-video and image-to-video at $0.075 a clip, or unlimited on the $149 plan. Wan 2.7 (see the Wan 2.7 API page at modelslab.com/wan-2-7) adds text-to-image and image edit in the same family and runs on Alibaba Cloud, so it is billed per call as a third-party model and is not covered by Open Source Unlimited. For every video model side by side, see the AI Video API page at modelslab.com/ai-video-api.

Yes. Pass a webhook URL and ModelsLab POSTs the completed MP4 URL to your endpoint when generation finishes.

No. ModelsLab is paid-only. Creating an account is free, but generating video needs a plan, from $21/month for 3,250 API calls.

Ready to create?

Start generating with Wan 2.2 API — Text-to-Video & Image-to-Video on ModelsLab.

Try Wan 2.2Wan 2.7 APIAI Video API