LeapTalk, a framework developed by researchers from Shanghai Jiao Tong University and Harbin Institute of Technology, achieves real-time talking-head generation with a single forward step, scaling to arbitrarily long videos. The method addresses the latency-quality trade-off in long-form generation by using a single-step bridge distillation scheme. Departing from conventional noise-to-data paradigms, LeapTalk introduces a data-to-data transport formulation based on a Brownian bridge, anchored by a persistent reference to mitigate identity drift and enhance long-term temporal stability. To enable knowledge transfer from a pre-trained diffusion teacher to the student bridge model, the framework employs a heterogeneous distillation approach with an SNR-aligned time transformation. Additionally, an audio-driven classifier-free guidance mechanism maintains fine-grained lip synchronization under extreme step reduction. LeapTalk demonstrates real-time streaming inference on a single H200 GPU, with only 1 inference step each chunk, achieving up to 200 FPS while preserving excellent video quality. Extensive experiments show the method achieves high-fidelity and temporally consistent video generation with only 1 step at up to 200 FPS, significantly outperforming existing approaches in both efficiency and stability.