Wan-Streamer v0.2 Higher Resolution, Same Latency

Jul 16, 2026

Upload by Maeve

Wan-Streamer v0.2 is a latency-preserving upgrade to the native-streaming end-to-end audio-visual interaction model. It keeps the v0.1 formulation while raising the interactive output stream from 192×336 to 640×368 at 25 FPS, maintaining about 200 ms model-side signal-to-signal latency. The higher-resolution stream supports scene-grounded mid-shot agents with legible posture, gaze, hands, nearby objects, and local layout during real-time conversation, using a single-GPU thinker path and a multi-GPU context-parallel performer group to sustain roughly 550 ms remote interaction latency under a 350 ms network budget.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages7

Reading Time18 min

Introduction

Native-streaming formulation in v0.1

Latency-preserving resolution upgrade in v0.2

Serving topology: thinker-performer split

Supported visual composition and target constraints

arXiv :2607 .04443v 3 [ cs .CV] 8 Jul 2026 Wan-Streamer v0.2: Higher Resolution, Same Latency Wan Team, Alibaba Group See Contributions and Acknowledgements for the full author list. Abstract We present Wan-Streamer v0.2 , a latency-preserving upgrade of the native-streaming, end-to-end audio-visual interaction model. v0.2 keeps the v0.1 modeling formulation, but raises the interactive output stream from 192 ×336 to 640×368 while preserving approximately 200 ms model-side signal-tosignal latency at 25 FPS. The higher-resolution stream supports scene-grounded mid-shot agents whose posture, gaze, hands, nearby objects, and local scene layout remain legible during real-time conversation. To support the larger visual stream without adding user-visible delay, v0.2 keeps the thinker as a single-GPU low-latency path for streaming perception, the short language/state Transformer pass that builds the generation cache, and final decoding. The performer becomes a multi-GPU Ulysses-style context-parallel group for the expensive next-unit latent generation. Each performer rank writes incoming K/V into a pre-sharded local cache. The long high-resolution latent video sequence is split across ranks for denoising and gathered through Ulysses communication, while the much shorter audio latent sequence is generated without sequence sharding. In this split, the thinker’s language/state computation reaches the performer only as K/V conditioning, so no separate language sequence has to be communicated inside the performer group. This concentrates additional hardware on visual generation while preserving the compact thinker-performer boundary, keeping total remote interaction latency at approximately 550 ms when a 350 ms bidirectional network budget is included. Website: [https://wan-streamer.com/](https://wan-streamer.com/) 1 Introduction Wan-Streamer v0.2 directly upgrades Wan-Streamer v0.1 [1] . Real-time audio-visual interaction sits at the intersection of full-duplex spoken dialogue, multimodal perception, streaming video generation, and interactive digital humans. Full-duplex speech systems show that natural dialogue should not be reduced to alternating ASR–LLM–TTS turns [2, 3] . Omni-modal models extend perception to image, video, and audio inputs [4, 5], while video generation and causal rollout methods provide the visual synthesis and streaming foundations needed for interactive output [6–8] . In parallel, real-time avatars and digital-human systems have advanced audio-driven faces, streaming visual agents, and end-to-end embodied interaction [9–12] . Wan-Streamer v0.1 established a native-streaming formulation for this setting: user and agent text, audio, and video are represented on one causal timeline and modeled by a single Transformer. Unlike cascaded visual-agent systems, this formulation keeps perception, response timing, speech, visible listening behavior, and synchronized video response inside one causal interaction state. It closes the audio-visual interaction loop, but the preliminary 192p output limits the visual range. Close-up video-call framing preserves facial response and speaking behavior, while wider compositions leave body posture, nearby objects, and scene context too compressed for scene-grounded interaction. Wan-Streamer v0.2 is a latency-preserving resolution upgrade. It raises the interactive output stream from Figure 1 Wan-Streamer remains one native-streaming model: language, audio, and video inputs and outputs are represented on a shared causal timeline and coordinated by block-causal attention. v0.2 keeps this formulation while increasing the output resolution and changing the deployment strategy described in Fig. 2. 192 ×336 to 640 ×368 at 25 FPS while keeping approximately 200 ms model-side response latency. This target is constrained by streaming causality rather than offline rendering quality: every 160 ms unit must process current user observations, update the shared interaction st

FAQs

What improvement does Wan-Streamer v0.2 bring over v0.1?

It upgrades the interactive output resolution from 192×336 to 640×368 while keeping the same real-time latency targets. The model side response remains around 200 ms at 25 FPS.

How does v0.2 preserve latency while increasing output resolution?

v0.2 keeps a single-GPU low-latency thinker path for streaming perception, language/state updates, KV-cache construction, and final causal decoding. It moves the high-cost 640×368 latent generation to a multi-GPU Ulysses-style context-parallel performer group.

What kinds of visual behavior does the higher-resolution stream enable?

It supports scene-grounded mid-shot agents where posture, gaze, hands, nearby objects, and local scene layout stay legible during real-time conversation, expanding beyond close-up framing.

Jul 16, 2026

Upload by Maeve