Enterprise Plans
AI Application Development
AI Translator API
AI SEO/GEO Service
Web Scraping Service
Log in
const main = async () => {
const response = await fetch('https://api.ai.cc/v2/video/generations', {
method: 'POST',
headers: {
Authorization: 'Bearer ',
'Content-Type': 'application/json',
},
body: JSON.stringify({
model: 'alibaba/wan2.2-14b-animate-replace',
prompt: 'Mona Lisa puts on glasses with her hands.',
video_url: 'https://storage.googleapis.com/falserverless/example_inputs/wan_animate_input_video.mp4',
image_url: 'https://s2-111386.kwimgs.com/bs2/mmu-aiplatform-temp/kling/20240620/1.jpeg',
resolution: "720p",
}),
}).then((res) => res.json());
console.log('Generation:', response);
};
main()
import requests
def main():
url = "https://api.ai.cc/v2/video/generations"
payload = {
"model": "alibaba/wan2.2-14b-animate-replace",
"prompt": "Mona Lisa puts on glasses with her hands.",
"video_url": "https://storage.googleapis.com/falserverless/example_inputs/wan_animate_input_video.mp4",
"image_url": "https://s2-111386.kwimgs.com/bs2/mmu-aiplatform-temp/kling/20240620/1.jpeg",
"resolution": "720p",
}
headers = {"Authorization": "Bearer ", "Content-Type": "application/json"}
response = requests.post(url, json=payload, headers=headers)
print("Generation:", response.json())
if __name__ == "__main__":
main()
-
AI Playground
Test all API models in the sandbox environment before you integrate.
We provide more than 300 models to integrate into your app.
Product Detail
The Wan 2.2 14B Animate Replace model represents a breakthrough in AI video generation, specifically engineered for precise character replacement within existing video footage. This advanced model excels at maintaining the original video's intrinsic elements – its scene, background, camera angles, and precise timing – while seamlessly substituting the original person with a new character derived from a single reference photo. Users have the flexibility to choose between replacing just the face or executing a full-body substitution, all while preserving the original body poses and ensuring perfectly synchronized lip movements.
Technical Specifications
- ✨ Model Size: 14 billion parameters powering the generation backbone.
- 🧠 Architecture: A robust Diffusion Transformer video generator, featuring a mixture-of-experts design for enhanced capacity and efficient compute.
- 🌌 Latent Space Processing: Utilizes a custom 3D causal Variational Autoencoder (VAE) (~127M parameters) for spatio-temporal latent video compression.
- 🕰️ Causality: Temporal causality ensures that future frames do not influence past frames, leading to stable and coherent motion generation.
- 👁️ Attention Mechanism: Pooled spatio-temporal self-attention across video frames and pixels.
- ✍️ Conditioning: Cross-attention to text features via a T5 encoder for optional text-driven control.
- 📥 Input: Requires a single reference image (for identity) and a reference video (for motion).
- 📤 Output: Produces a video with the replaced character, delivered in 720p resolution at 24 frames per second.
Performance Benchmarks
- 🖼️ Video Quality: Delivers high-fidelity character replacement with remarkably smooth motion and natural facial expressions.
- 📏 Resolution & Frame Rate: Consistently supports 720p resolution at 24 frames per second.
- ⚡ Latency: Local generation speed is GPU-dependent; H100 GPUs offer significantly faster inference compared to consumer-grade GPUs.
- 📉 Resource Efficiency: The innovative mixture-of-experts architecture boosts model capacity without a proportional increase in computational cost.
Key Features
- 👤 Character Replacement: Effortlessly swap the original person in a video with a new one using just a single reference image.
- 🎭 Full or Partial Replacement: Choose the desired level of replacement – from just the face to a complete full-body substitution.
- 🕺 Pose & Expression Preservation: Maintains the integrity of original body poses, head movements, and lip synchronization for incredibly natural animation.
- 🏞️ Scene Consistency: Guarantees that the background, camera angles, lighting, and timing remain perfectly intact.
- 🌟 High Realism: Leverages skeleton-based motion tracking and fine facial encoding to produce smooth, highly realistic animations.
- 💻 Local Deployment: Capable of running locally with appropriate hardware setups, supporting high-quality output generation.
API Pricing
- 480p: $0.042
- 580p: $0.063
- 720p: $0.084
Use Cases
- 📈 Marketing & Advertising: Dynamic video character replacement for engaging marketing and advertising content.
- ✨ Virtual Influencers: Creation of virtual influencers and avatars with realistic, real-time expression mimicking.
- 🎬 Film & Video Production: Streamline film and video pre-visualization and reshoots without the need for new filming.
- 🎨 Personalized Content: Generate personalized user-generated content featuring custom characters.
- 📱 Social Media Animation: Animate photos for captivating social media posts and entertainment.
- 📚 Educational Content: Customize educational and training videos with tailored characters.
- 🔒 Privacy Preservation: Enhance privacy by replacing faces in existing footage for sensitive content creation.
- 🎭 Digital Effects: Facilitate digital effects and deepfake production, with a focus on ethical controls.
Code Sample
<snippet data-name="alibaba.create-video-to-video-generation" data-model="alibaba/wan2.2-14b-animate-replace"></snippet>
Comparison with Other Models
vs Stable Diffusion Video: Wan 2.2 Animate Replace primarily focuses on end-to-end character replacement in videos, offering holistic expression and motion transfer. This capability surpasses standard Stable Diffusion extensions, which are typically limited to short clip generation and often exhibit less consistent temporal control. Furthermore, Wan 2.2 can process longer videos (up to several minutes), a significant advantage over the typically shorter outputs from Stable Diffusion video models.
vs Imagen Video (Google): Imagen Video is largely geared towards generating videos from text prompts with high visual quality, but it does not inherently offer specific character replacement features. Wan 2.2 Animate Replace's unique selling point lies in unifying animation and replacement modes, providing detailed control over expressions and motion, thereby catering specifically to character-centric workflows.
vs Meta Make-A-Video: While Make-A-Video generates short video clips from text, it lacks the targeted character substitution that Wan 2.2 Animate Replace specializes in. Wan 2.2 excels in character replacement with precise synchronization of pose and lips within existing videos, making it more practical for post-production and video editing workflows, contrasting with Make-A-Video's focus on general scene creation.
API Integration
Wan 2.2 Animate Replace is accessible via the AI/ML API. Comprehensive documentation is available here.
Frequently Asked Questions (FAQ)
What is Wan 2.2 14B Animate Replace?
Wan 2.2 14B Animate Replace is an advanced AI video generation model designed to accurately replace characters in existing videos. It intelligently maintains the original video's scene, background, camera angles, and timing, while integrating a new character from a reference image.
Can I replace only the face or the entire body?
Yes, the model offers flexibility. You can choose to replace just the face of the original character or opt for a full-body replacement, with both options preserving original body poses and lip synchronization for natural animation.
How does Wan 2.2 ensure video consistency after replacement?
The model is engineered to maintain scene consistency by keeping the original background, camera angles, lighting conditions, and timing fully intact. It focuses solely on replacing the character while preserving all other elements of the video.
What are the primary applications of this character replacement technology?
Key applications include video character replacement for advertising, creating virtual influencers and avatars, film pre-visualization, personalized user-generated content, educational video customization, and privacy-preserving content creation.
Is local deployment an option for Wan 2.2 Animate Replace?
Yes, Wan 2.2 14B Animate Replace can be deployed and run locally, provided you have the appropriate hardware setup. This allows for high-quality output generation directly on your own infrastructure.
Learn how you can transformyour company with AICC APIs
