Kunlun Wanyi announced that its TianGong AI large model SkyReels V4 has ranked first globally in the text-to-video (with audio) category of Artificial Analysis. The model's performance significantly surpasses mainstream models such as Kling3.0, Google Veo3.1, Vidu Q3, and OpenAI Sora2, becoming the AI large model with the strongest video generation capability in the world today.

image.png

Core Breakthroughs: Full-modal Reinforcement Learning and Logical Reasoning

SkyReels V4 has achieved two core technological transformations in its architecture, solving the issues of consistency and narrative logic in video generation:

  • Reinforcement Learning System (RL): By building a full-modal semantic Reward model and adopting a step-by-step curriculum learning path, the model is infused with logical reasoning capabilities, achieving commercial-grade long-sequence generation of 1080p for 15 seconds.

  • Advanced Reference Tasks: Added "keyframe reference" and "grid map reference" capabilities. The former can accurately infer coherent scenes between nodes, while the latter allows uploading multiple story images to ensure consistent character features and scene styles throughout short film creation.

With the ranking at the top of the list, the API entry of SkyReels V4 is now officially open to all scenarios. Its capabilities fully cover all core functions of the model:

  • Full Function Coverage: Including text-to-video, image-to-video, multimodal reference generation, video editing and repair, as well as audio-visual joint generation.

  • Low-Threshold Empowerment: E-commerce, education, content platforms, and development teams can directly call the world's leading audio-visual generation capabilities without investing significant R&D costs.

Kunlun Wanyi has previously released and open-sourced multiple models in the SkyReels series. From V1’s human-driven generation to V2’s long-form video generation, and now V4's comprehensive breakthrough in audio-visual synchronization and logical expression, SkyReels demonstrates a transition from "being able to generate" to "generating well."

Currently, the technical report of SkyReels V4 has been released simultaneously. Developers can obtain the API documentation and carry out business integration through its official platform. This development marks that China's AI is now in a global leading position in the vertical sector of audio-visual content generation.