Next-generation platform delivers industry-leading improvements in motion accuracy, color fidelity, and visual quality
Hong Kong, China–(Newsfile Corp. – November 23, 2025) – SoulGen today announced the launch of SoulGen 2.0, a major upgrade to its image-to-video AI generation platform. This establishes new industry benchmarks across motion accuracy, color reproduction, and semantic understanding.
To view an expanded version of this graphic, please visit:
https://images.newsfilecorp.com/files/11790/275624_930aa79a24eb482c_001full.jpg
This release addresses key technical challenges that have previously limited the quality of AI video generation, including body pose distortion, color mismatch, and texture degradation.
Significant performance improvements across core functionality
SoulGen 2.0 shows significant performance improvements across all three major generation modes.
Image to video (I2V) The generation achieves a 23% performance improvement and enables more accurate transitions from static images to dynamic video content.
Text to video (T2V) With a 17% increase in functionality, users can now create high-quality videos directly from text descriptions.
Video extension These features show a 19% performance increase and provide a more seamless and natural video continuity.
These enhancements translate directly into production-ready output suitable for professional creative workflows and demanding commercial applications.
Revolutionary human movement precision
One of the most important advances in SoulGen 2.0 focuses on human pose accuracy. The platform achieves a mean position error per joint (MPJPE) of 42.3 mm, which represents a 38.2% reduction from 68.5 mm in the previous version.
This improvement improves the position of each joint in the body by more than 2.6 centimeters, effectively eliminating the noticeable pose distortions that plagued early AI video generation systems.
The percentage of correct keypoints (PCK@150mm) index reaches 96.8%, indicating that almost all keypoints are detected accurately, ensuring anatomically correct and natural movements.
Industry-leading color reproduction
Color accuracy is another area of notable achievement. SoulGen 2.0 has a ΔE2000 score of 1.27, representing a 73.7% improvement from version 1.0’s score of 4.82.
This achievement is particularly important because the color difference falls below a critical threshold of 2.0, at which point the human eye no longer perceives the change.
Content creators can now leverage generated videos to maintain faithful color matching with reference images, eliminating noticeable color shifts that previously required extensive post-processing corrections.
Excellent detail and texture preservation
Improvements in texture quality are demonstrated across multiple technical metrics. The Structural Similarity Index (SSIM) score improved to 0.947, closer to the excellence benchmark of 0.95.
Meanwhile, the Learned Perceptual Image Patch Similarity (LPIPS) metric decreased by 56% to 0.096.
These measurements confirm that SoulGen 2.0 retains fine detail and texture at nearly the same level as the base model, directly addressing the smoothing and loss of detail commonly seen in AI-generated content.
The peak signal-to-noise ratio (PSNR) rating is 41.8dB (28.2% increase), further proving the platform’s entry into the realm of high-quality image generation.
Enhanced motion naturalness and semantic consistency
Motion naturalness has received innovative enhancements, and the Fréchet Inception Distance (FID) score has been reduced from 8.45 to 2.73, a 67.7% reduction, positioning SoulGen 2.0 as one of the industry’s leading AI video generation platforms.
This metric measures how similar the generated content is to real-world video delivery and directly impacts perceived realism and fluidity.
Semantic understanding also showed meaningful progress, with CLIP scores increasing by 18.5% to 0.891.
This improvement reflects significantly greater coordination between text prompts and visual output, allowing for more intuitive creative control and reducing iterative adjustment cycles.
Production-ready quality for professional applications
The technical improvements in SoulGen 2.0 deliver tangible operational benefits for content creators and production teams.
Reduced body distortion minimizes costly retakes and manual corrections. Accurate color reproduction eliminates the need for time-consuming color grading sessions and streamlines your workflow.
Enhanced texture preservation provides professional-grade output suitable for commercial deployment without additional polishing.
These advancements position SoulGen 2.0 as a production-ready solution that bridges the gap between AI-assisted generation and professional quality standards.
Comprehensive performance metrics
The following table details the quantitative improvements achieved in SoulGen 2.0 compared to version 1.0.
| Evaluation index | Soulgen 1.0 | Soulgen 2.0 | improvement | What this means for video quality |
| MPJPE (mm) | 68.5 | 42.3 | ↓38.2% | Body movements are natural and precise, with no unnatural distortions. |
| PCK@150mm (%) | 87.3 | 96.8 | ↑10.9% | Almost all body parts are correctly positioned and “broken limb” artifacts are eliminated. |
| ΔE2000 | 4.82 | 1.27 | ↓73.7% | Perfect color accuracy – no fading or tonal inconsistencies. |
| SSIM | 0.823 | 0.947 | ↑15.1% | Details such as facial features and textures remain crisp and clear. |
| PSNR (dB) | 32.6 | 41.8 | ↑28.2% | Broadcast-quality clarity with minimal noise and compression artifacts. |
| LPIPS | 0.218 | 0.096 | ↓56.0% | The video looks real and authentic to human perception. |
| F.I.D. | 8.45 | 2.73 | ↓67.7% | Natural, smooth movements without any jerky or robotic movements. |
| clip score | 0.752 | 0.891 | ↑18.5% | The AI interprets text prompts and reference images accurately on the first try. |
availability
SoulGen 2.0 is now available to all users and represents a significant milestone in AI video generation technology and continued advancement in accessible, professional-quality creative tools.
About Soulgen
SoulGen is an AI-powered video generation platform dedicated to providing professional quality content creation tools. The platform specializes in image-to-video generation, text-to-video creation, and video enhancements, serving creative professionals and production teams around the world.
Contact details:
Representative name: Sean Banks
Representative email address: business_shawnbanks@soulgen.ai
Email: business_shawnbanks@soulgen.ai
Phone/WhatsApp: 852 68452818
Company name: Wavedance Intelensic
Address: Hysan Avenue, Causeway Bay, Hong Kong.
Website/URL: https://www.soulgen.ai/
Social link: https://x.com/SoulGen_AI
To view the source version of this press release, please visit https://www.newsfilecorp.com/release/275624.
