Kling: Kling v3 Text to Video API

Kling Video 3.0 is a frontier-level Spatio-Temporal Foundation Model that incorporates the "Physical Reasoning Engine" (PRE) to solve long-standing AI video artifacts. Unlike previous DiT-based models, v3.0 performs a preliminary "logical sketch" of physical interactions (gravity, momentum, light bounce) before pixel synthesis. It supports "Zero-Shot Character Consistency," allowing users to maintain the same protagonist across multiple 60-second clips without manual fine-tuning. In 2026, it is the primary engine for high-end digital marketing and indie film production, offering a "Director Mode" for precise, natural-language-based camera and lighting control.

• Input: text, video
• Output: video
• File input: Supported
• Released: 2026-01-30
• Knowledge cutoff: 2026-02-28