Wan Video Local Cloud

Wan 2.2 T2V

Generate high-quality video clips from text descriptions with Wan 2.2, one of the most capable open-source video generation models.

$0.15

per video in the cloud

$0

on your own GPU

Run it free Use it in Claude or the API

Overview

Wan 2.2 T2V (Text-to-Video) generates dynamic video clips directly from text prompts. It is one of the strongest open-source video generation models available, producing coherent motion, detailed scenes, and consistent subjects across frames.

Key Features

High Motion Quality: Generates smooth, temporally consistent video with natural physics and movement
Scene Complexity: Handles multi-element scenes with backgrounds, characters, lighting, and environmental detail
Strong Prompt Adherence: Faithfully interprets detailed text descriptions into visual sequences
Open Source: Based on the open Wan 2.2 framework, offering transparency and reliability

Best For

Creative video generation from scratch
Scene and concept visualization
Short cinematic clips and mood videos
Storyboarding and rapid video prototyping

Prompt Tips

Describe the scene, action, camera movement, and mood for best results
Specify lighting conditions and environment (e.g. "golden hour", "rainy street", "studio lighting")
Clear, action-focused prompts produce the most coherent motion

Use it

Three ways to run Wan 2.2 T2V

Download Imference Desktop and open it.
In Local models, pick Wan 2.2 T2V. It downloads once.
Write a prompt and press Generate. Free, on your GPU.

Get the app

Make a video with Wan 2.2 T2V ("wan22-t2v"): a paper boat drifting down a rainy street, slow push in

curl https://imference.com/generate \
  -H "Authorization: Bearer $IMFERENCE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "wan22-t2v",
       "format_code": "wan-square",
       "prompt": "a paper boat drifting down a rainy street, slow push in"}'

Claude needs the MCP server. API details in the reference.

Specs