Paper 2601.10103
FlowAct-R1: Towards Interactive Humanoid Video Generation
- Published
- Jan 2026
- Research lab
- Independent
- Citations
- 7
- GitHub
- Not linked
01 In brief
Summary
FlowAct-R1 is a framework for real-time interactive humanoid video generation, built on a MMDiT architecture (Seedance).
It enables streaming synthesis of arbitrary-length videos with low latency, achieving 25fps at 480p resolution with a time-to-first-frame of about 1.5 seconds.
The method uses a chunkwise diffusion forcing strategy with a self-forcing variant to reduce error accumulation and maintain long-term temporal consistency.
It supports holistic full-body control via audio and text, allowing natural transitions between behavioral states.
The framework includes a structured memory bank (reference, long-term, short-term, and denoising stream) and a three-stage training curriculum (autoregressive adaptation, joint audio-motion training, and distillation).
Distillation reduces the denoising process to 3 NFEs, and system-level optimizations (FP8 quantization, frame-level hybrid parallelism, operator fusion, asynchronous VAE decoding) enable real-time performance.
A Multimodal Large Language Model (MLLM) is integrated for action planning to enhance behavioral naturalness.
User studies show FlowAct-R1 outperforms KlingAvatar 2.0, LiveAvatar, and Omnihuman-1.5 in motion naturalness, lip-sync accuracy, frame stability, and motion richness.
The method generalizes across diverse character styles from a single reference image.
Ethical considerations include access control and use of AI-generated human images in demos.
02 From the paper
Abstract
Interactive humanoid video generation aims to synthesize lifelike visual agents that can engage with humans through continuous and responsive video. Despite recent advances in video synthesis, existing methods often grapple with the trade-off between high-fidelity synthesis and real-time interaction requirements. In this paper, we propose FlowAct-R1, a framework specifically designed for real-time interactive humanoid video generation. Built upon a MMDiT architecture, FlowAct-R1 enables the streaming synthesis of video with arbitrary durations while maintaining low-latency responsiveness. We introduce a chunkwise diffusion forcing strategy, complemented by a novel self-forcing variant, to alleviate error accumulation and ensure long-term temporal consistency during continuous interaction. By leveraging efficient distillation and system-level optimizations, our framework achieves a stable 25fps at 480p resolution with a time-to-first-frame (TTFF) of only around 1.5 seconds. The proposed method provides holistic and fine-grained full-body control, enabling the agent to transition naturally between diverse behavioral states in interactive scenarios. Experimental results demonstrate that FlowAct-R1 achieves exceptional behavioral vividness and perceptual realism, while maintaining robust generalization across diverse character styles.