BEIJING, July 14, 2026 (GLOBE NEWSWIRE) -- Cheer Holding, Inc. (NASDAQ: CHR) ("Cheer Holding" or the "Company"), a leading provider of advanced mobile internet infrastructure and platform services, today announced that its proprietary AI portrait and digital identity creation platform, Klon AI, has exited invite-only beta testing and is now fully available to global users on the App Store and Google Play.
Following successful closed beta testing in North America, Latin America, Japan, South Korea and Southeast Asia, Klon AI stands out as one of the few consumer-grade applications to deeply integrate AI visual generation with personal digital identity management.
Product Features and Highlights
Users can generate professional studio-quality portraits in multiple styles and build a sustainable, evolving digital twin simply by uploading 3–5 portrait photos — no professional equipment or photography team required. The official version introduces four core modules:
Multi-style portrait generation engine
Global scene library with thousands of templates
Cross-scene identity-consistent digital twin system
Intelligent tool that automatically converts static portraits into social short videos
Together, these form a complete creation-to-distribution loop. The platform's key differentiator is its strong cross-scene identity consistency, ensuring users' facial features and personal style remain unified across portraits, short videos, virtual avatars, and other formats — creating recognizable personal digital assets.
Technological Leadership and Advantages
Rather than pursuing unstable frontier concepts, Klon AI focuses on deep engineering implementation of proven, scalable technologies:
High-Quality Generation: Built on latent diffusion models with DiT architecture for enhanced detail, skin texture, and compositional stability.
Identity Consistency: Combines Identity Consistency algorithms with InstantID/IP-Adapter-FaceID training-free identity injection and lightweight LoRA fine-tuning, validated by ArcFace embeddings, enabling second-level creation of personalized digital twins.
Composition and Image Quality: ControlNet for precise pose and composition control, paired with mature face restoration and super-resolution techniques.
Static-to-Video Conversion: Mature image-to-video diffusion models that maintain identity consistency while generating smooth short videos.
Speed and Cost Efficiency: LCM/Turbo distillation acceleration, model quantization, and elastic cloud inference reduce single-image generation latency to seconds. Internal testing shows identity feature retention exceeding 95% across continuous creations.