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-vace-fun-a14b-inpainting',
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-vace-fun-a14b-inpainting",
"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
Overview: Advanced Video Inpainting with Wan 2.2 Vace
Wan 2.2 Vace Inpainting is a cutting-edge video-to-video generative AI model engineered for high-quality inpainting within video content. This powerful tool revolutionizes video editing by allowing users to seamlessly mask and modify specific regions in videos. It meticulously maintains contextual continuity, motion consistency, and precise detail restoration, ensuring a natural and polished final output. Leveraging advanced multimodal understanding and adaptive video generation technologies, Wan 2.2 Vace is optimized to deliver impressive results at resolutions up to 720p.
Technical Specifications & Performance
⚙️ Core Architecture
- Model Architecture: Multimodal video and image transformer backbone with adaptive scene and motion prediction.
- Parameter Size: 14 billion parameters, ensuring fine-grained detail in video synthesis.
- Resolution Range: Outputs up to 720p; default usage often at 480p for optimized speed.
- Frame Rate Processing: Operates at 16 frames per second for stable video synthesis.
- Input/Output Formats: Supports MP4, MOV, WebM, M4V, GIF for video; JPG, JPEG, PNG, WebP, GIF, AVIF for images.
- Memory Use: Efficient GPU usage, supporting both local and cloud-based workflows with a moderate memory footprint.
📊 Performance Benchmarks
- Inpainting Quality: Excels in preserving context and texture detail during localized video edits.
- Temporal Stability: Strong performance in motion continuity, minimizing flickering compared to baseline models.
- Resolution Scaling: High fidelity at 720p with minor speed trade-offs; faster processing at 480p with maintained visual consistency.
- Maximum Clip Length: Reliably supports up to 80-81 frames; quality may degrade beyond this.
Key Features for Seamless Video Editing
- ✨ Video Inpainting with Masking: Enables selective editing by using a source video and a corresponding mask video to inpaint or replace designated areas.
- ✅ Temporal Consistency: Guarantees smooth motion flow and coherence across frames, preventing common flickering or artifacts during inpainting.
- 🔬 High Detail Restoration: Reconstructs fine textures and intricate details within the masked region for a highly natural appearance.
- 📺 Resolution Support: Offers flexible video output at 480p, 580p, and up to 720p, with quality scaling based on resource availability.
- 📥 Flexible Input Types: Accepts various video formats (MP4, MOV, WebM, M4V, GIF) and image formats (JPG, PNG, WebP) for mask or reference inputs.
- 🤝 Integration with ComfyUI: Compatible with ComfyUI workflows, allowing combination with text-to-video, image animation, and outpainting pipelines.
API Pricing
- 360p: $0.0525
- 540p: $0.07875
- 720p: $0.105
Versatile Usage Scenarios
- 🎬 Professional Video Post-production: Ideal for VFX touch-ups, object removal, and intricate scene re-editing.
- 📈 Digital Marketing Content Creation: Facilitates automated video personalization and brand adaptation for engaging campaigns.
- 🎓 Educational Video Material Enhancement: Enables visual reconstructions or dynamic content update animations.
- 🎨 Creative Arts & Digital Storytelling: Supports seamless animation replacements and complex visual effects.
Code Sample
<snippet data-name="alibaba.create-video-to-video-generation" data-model="alibaba/wan2.2-vace-fun-a14b-inpainting"></snippet>
Comparison with Other Leading Models
Wan 2.2 Vace vs. Seedance 1.0: While Seedance 1.0 offers a slightly higher frame rate output at 24 fps (beneficial for ultra-smooth video generation), Wan 2.2 Vace provides superior integration for inpainting flexibility and operates efficiently on more modest hardware, making it broadly accessible to creators.
Wan 2.2 Vace vs. Veo 3: Veo 3 excels in advanced closed-source video enhancement with higher resolution and speed, though at a significantly higher cost. Wan 2.2 Vace stands as a strong open-source alternative, offering robust multimodal inpainting capabilities and easier API integration, suitable for a wide range of professional uses.
Wan 2.2 Vace vs. Generic Baseline Video Inpainting: Compared to basic baseline models, Wan 2.2 Vace delivers significantly improved texture detail restoration and motion coherence. Baseline models frequently suffer from more artifacts and flickering, positioning Wan 2.2 Vace as the preferred choice for high-quality video inpainting projects.
API Integration
Access Wan 2.2 Vace Inpainting via AI/ML API. Comprehensive documentation is available here.
Frequently Asked Questions (FAQ)
Q1: What is Wan 2.2 Vace Inpainting primarily designed for?
A1: Wan 2.2 Vace Inpainting is primarily designed for high-quality video-to-video inpainting tasks, enabling users to mask, modify, and seamlessly fill specific regions in videos while preserving context and motion continuity.
Q2: What resolution does Wan 2.2 Vace Inpainting support?
A2: The model supports output resolutions up to 720p, with default usage often set at 480p for faster processing while maintaining visual consistency.
Q3: How does Wan 2.2 Vace ensure temporal consistency in edited videos?
A3: It integrates a multimodal video and image transformer backbone with adaptive scene and motion prediction, specifically designed to maintain smooth motion flow and coherence across frames, preventing flickering or artifacts.
Q4: Can Wan 2.2 Vace Inpainting be integrated with other AI tools?
A4: Yes, it is compatible with ComfyUI workflows, allowing users to combine inpainting tasks with other pipelines like text-to-video generation, image animation, and outpainting.
Q5: What are the main advantages of Wan 2.2 Vace over generic inpainting models?
A5: Wan 2.2 Vace offers significantly improved texture detail restoration and superior motion coherence, leading to fewer artifacts and flickering compared to basic baseline models, making it ideal for quality-critical projects.
Learn how you can transformyour company with AICC APIs
