Wan 2.1Wan 2.1

Wan 2.2: Free AI Video Generator – Create Cinematic Videos

Wan 2.2, the open-source AI video model for stunning text-to-video and image-to-video generation. Cinematic quality, realistic motion, and precise control!

Input Section

Credits: 0

Reference Image (Optional)

Click to select Image

Or drag Image here

Prompt

Mode

Duration (seconds)

Aspect Ratio

Tips: A failed generation won't cost a credit.

Output Section

Please login to view your generation history.

Natural Sunlight Scene

Soft natural sunlight, warm tones, side lighting. A young girl sits in a tall grassy field under clear blue skies. Two fluffy donkeys stand behind her. Around 11 or 12, she wears a simple floral skirt, her hair tied in cotton pigtails. She smiles gently while playing with wildflowers. The sun bathes the countryside in warmth, creating a peaceful, innocent vibe. Shot from the side, mid-close-up, with low contrast and gentle highlights.

Indoor Artificial Light

Soft artificial light, warm tones, indoor side lighting. In a dim room, a foreign girl in a blue plaid shirt sits near a vintage lamp. Her ponytail is neat, with a few strands adding softness. She looks slightly to the side, lips parted as if listening. A desk lamp glows from the side, highlighting her face and silhouette. A dark cabinet in the background adds depth and mystery. Shot over-the-shoulder, mid-close-up, low saturation, intimate framing.

Moonlit Room Scene

Moonlight from outside window, soft and cool. A young woman stands quietly in an old-style room with tiled walls and wooden furniture. She has short black hair, sharp eyebrows, and deep blue eyes reflecting the moonlight. Wearing a black dress with a white collar, she stares out the window with a calm, restrained expression. Moonlight filters through, casting soft shadows and a serene atmosphere. The camera focuses tightly on her face and posture, evoking quiet emotional depth.

Bedside Lamp Scene

Warm bedside lamp, ambient interior lighting. In low indoor light, a white foreign woman sits on the edge of a bed wearing a printed white T-shirt. Her hair drapes over her shoulders, and a necklace hangs from her neck. A soft bedside lamp glows in front of her, illuminating her features. In the background, another woman dressed in black is watching her. Captured in mid-close-up, with a grounded, realistic tone.

Fireplace Scene

Firelight glow, flickering and warm. A man in a white shirt and brown vest stands in front of a fireplace. Captured from an over-the-shoulder angle, he gazes at someone off to the right edge of the frame. The fire casts soft, moving highlights across his face and the surroundings, emphasizing warmth and intimacy.

School Hallway Scene

Cold fluorescent light and sidelight from hallway windows. A young woman stands in a narrow hallway beside blue metal lockers. Her long black hair and plaid shirt contrast against the industrial background. Her calm expression turns subtly anxious as her eyes drift toward the camera. Light from ceiling tubes and side windows casts cool, soft shadows, enhancing a quiet and tense atmosphere.

Outdoor Urban Scene

Overcast daylight, diffused and cool. A foreign man stands outside under a gray sky, dressed in black and gray layered clothes. His eyes meet the lens seriously. Brown buildings and lit windows form the backdrop. The camera slowly moves forward. Black doors and blurred foreground objects enhance a moody and cinematic feel. Low contrast, single-subject composition.

Projection Room Scene

Mixed light from projection and LED effects, colorful reflections. Inside a dark room, a foreign man in a white tank top stands in front of a projector screen. He wears silver earrings and gazes into the distance with a contemplative expression. Blue and purple laser lines cross the background, casting shifting colors on his face. The mood is mysterious and dreamlike, with strong visual depth.

Kitchen Golden Hour

Golden-hour sunlight through the window, warm side light. In a cozy kitchen, a foreign man prepares food near a table. He wears a white shirt and black tie, scooping sugar from a jar into a blue cup. The warm evening light pours in through the window, touching the wallpaper, utensils, and his gentle gestures. Everything feels relaxed, real, and familiar.

What is Wan 2.2?

Wan 2.2 is a cutting-edge, new-generation multimodal generative AI model developed by WAN AI. It's designed to empower creators, artists, and businesses to produce breathtaking video content directly from text descriptions or still images. More than just a video generator, Wan 2.2 is an advanced platform that leverages sophisticated AI to deliver cinematic quality, realistic motion, and precise control over your output.

Features & Advantages of Wan 2.2

Wan 2.2 stands out from the crowd with its powerful features and distinct advantages:

Core Functionalities:

  • Text-to-Video (T2V) Generation: Simply describe your desired scene in text, and Wan 2.2 will bring it to life as a dynamic video.
  • Image-to-Video (I2V) Generation: Transform static images into captivating video clips, adding motion and dynamism to your visuals.
  • Hybrid Text-Image-to-Video (TI2V): Combine the power of text prompts with an initial image to guide your video generation for even more precise results.

How to Use Wan 2.2

Wan 2.2 lets you generate videos by providing clear, descriptive prompts. The more detailed your instructions, the better the video will match your vision. You can create videos in a few ways:

  1. Text-to-Video (T2V): Just type out what you want to see. For example, a basic prompt might be: "A cat, in a garden, running."
  2. Image-to-Video (I2V): Start with a still image and let Wan 2.2 bring it to life with motion.
  3. Hybrid Text-Image-to-Video (TI2V): Combine a text description with an initial image for more precise control over the generated video.

For advanced video creation, you can add more detail to your prompts:

  • Subject Description: Be specific about your main character or object (e.g., "a Tibetan dancer in ethnic costume").
  • Scene Description: Detail the environment (e.g., "a lush, ancient forest with dappled sunlight").
  • Motion Description: Specify how an action happens (e.g., "slowly waving her skirt").
  • Aesthetic Control: Define visual elements like lighting, camera angles, and shot types (e.g., "golden hour lighting, close-up shot").
  • Stylization: Set the overall mood or artistic style (e.g., "cyberpunk style," "dreamlike aesthetic").

When using Image-to-Video, your prompt should focus on:

  • Motion: Describe the actions of elements within your image (e.g., "person running quickly").
  • Camera Movement: Specify desired camera actions like panning, zooming, or keeping the camera fixed (e.g., "slow rotation of the camera," "fixed shot").

Key Advantages & Differentiators of Wan 2.2:

  • Cinematic-Level Aesthetic Control: Achieve unparalleled control over lighting, color, and composition, allowing you to craft videos with a truly professional and artistic touch.
  • Large-Scale Complex Motion Generation: Generate incredibly smooth and intricate movements within your videos, capturing the nuances of realistic action and dynamic scenes.
  • Precise Semantic Compliance: Wan 2.2 excels at understanding and accurately depicting complex scenes and multiple objects within your prompts, ensuring your generated videos align perfectly with your vision.
  • Innovative MoE Architecture: Utilizing a unique Mixture of Experts (MoE) architecture, Wan 2.2 employs specialized high-noise and low-noise expert models that adapt during the denoising process based on the Signal-to-Noise Ratio (SNR). This sophisticated design contributes to its superior generation quality and efficiency.
  • Exceptional Performance:
    • Speed: It is optimized for speed, capable of generating 5-second, 720P@24fps videos in under 9 minutes on consumer-grade GPUs like the NVIDIA RTX 4090, making it one of the fastest models in its class for this resolution.
    • Efficiency: Features an efficient compression ratio (e.g., 16x16x4 and 4x32x32 with patchification for the 5B model), optimizing resource usage.
    • Leading Benchmarks: On internal benchmarks like Wan-Bench 2.0, Wan 2.2-T2V-A14B has demonstrated leading performance in 5 out of 6 categories, outperforming commercial models such as Kling 2.0, Sora, and Seedance in key aspects like Dynamic Degree, Text Rendering, and Object Accuracy.
  • Open-Source & Accessible: Released under the Apache 2.0 license, Wan 2.2 supports commercial use and offers broad accessibility. It's integrated with popular platforms like ComfyUI and Diffusers, and available via API on WaveSpeed AI.
  • Prompt Extension & LoRA Training: Advanced features for fine-tuning and customizing your video generation, giving you even greater creative freedom.
  • Versatile Output Resolutions: Supports a wide range of output resolutions, including 1920x1080, 1080x1920, and 1440x1440 for T2V, and 1080P and 480P for I2V, catering to various project needs.

How to Install Wan 2.2 (Open Source)

As an open-source project, Wan 2.2's core inference code and model weights are publicly available. Here's a general guide to get started:

  1. Visit the Official GitHub Repository: The primary source for installation instructions and model weights is the Wan 2.2 GitHub repository.
  2. Review System Requirements: Check the README.md file on the GitHub page for detailed system requirements, including recommended GPU, memory, and software dependencies (e.g., Python version, PyTorch).
  3. Clone the Repository: Use Git to clone the repository to your local machine:
    git clone [https://github.com/Wan-Video/Wan2.2.git](https://github.com/Wan-Video/Wan2.2.git)
cd Wan2.2
  4. Install Dependencies: Install the required Python packages, typically listed in a requirements.txt file within the repository:
    pip install -r requirements.txt
  5. Download Model Weights: Follow the instructions on the GitHub page to download the pre-trained model weights (e.g., Wan2.2-TI2V-5B, Wan2.2-I2V-A14B, Wan2.2-T2V-A14B). These are crucial for the model to function.
  6. Run Inference: The repository will provide example scripts or notebooks demonstrating how to run the model for text-to-video or image-to-video generation.
  7. ComfyUI/Diffusers Integration: If you prefer a more user-friendly interface, look for instructions on integrating Wan 2.2 with ComfyUI or Hugging Face Diffusers, as these platforms often provide graphical interfaces for running AI models.

Always refer to the official GitHub repository for the most up-to-date and precise installation instructions.

10 FAQs about Wan 2.2

Here are some frequently asked questions about Wan 2.2:

  1. What types of videos can Wan 2.2 generate? Wan 2.2 can generate videos from text prompts (Text-to-Video), from still images (Image-to-Video), and from a combination of both (Text-Image-to-Video). It excels at creating complex scenes with detailed motion and semantic accuracy.

  2. Is Wan 2.2 free to use? Yes, Wan 2.2 is open-source under the Apache 2.0 license, which means it is free to use, including for commercial purposes.

  3. What kind of hardware do I need to run Wan 2.2? While it can run on consumer-grade GPUs, a powerful card like the NVIDIA RTX 4090 is recommended for optimal performance and faster generation times, especially for 720P videos.

  4. How does Wan 2.2 compare to Sora or Kling? Wan 2.2 has demonstrated leading performance in several categories on its internal Wan-Bench 2.0 benchmark, outperforming commercial models like Sora, Kling 2.0, and Seedance in aspects such as Dynamic Degree, Text Rendering, and Object Accuracy.

  5. Can I use Wan 2.2 for professional projects? Absolutely! With its Apache 2.0 license, cinematic control, and high-quality output, Wan 2.2 is suitable for a wide range of professional applications in content creation, advertising, animation, and more.

  6. What is the "MoE architecture" in Wan 2.2? MoE stands for Mixture of Experts. It's an innovative AI architecture where the model uses different "expert" sub-models (high-noise and low-noise) that specialize in different parts of the video generation process, particularly during denoising. This allows for more efficient and higher-quality results.

  7. What resolutions can Wan 2.2 generate? Wan 2.2 supports various resolutions, including up to 1920x1080, 1080x1920, and 1440x1440 for T2V, and 1080P and 480P for I2V, depending on the specific model version and your hardware.

  8. Where can I access Wan 2.2 if I don't want to install it locally? You can access Wan 2.2 through its API available on platforms like WaveSpeed AI, which allows you to use its capabilities without local installation.

  9. Can I fine-tune Wan 2.2 with my own data? Yes, Wan 2.2 supports advanced features like LoRA training, enabling users to fine-tune the model with specific datasets for customized video generation.

  10. What are the latest updates for Wan 2.2? As of July 2025, Wan 2.2 has seen recent updates including integration with ComfyUI and Diffusers, alongside the release of its inference code and model weights, making it more accessible to the community.