Text chat feels too flat
Most AI companions can write back quickly, but they cannot look present, react with visible listening behavior, or make a conversation feel face to face.
Wan Streamer AI Video Companion
Wan Streamer is a prelaunch real-time AI video companion for face-to-face conversations with an AI that can see, hear, understand, and respond with synchronized voice, expression, and motion.
Low-latency replies
Feels like a live call
~200ms model-side
Smooth video
Expression stays continuous
25fps
Clearer image
More readable face and scene
v0.2: 640×368
AI chat needs a face-to-face layer
People do not communicate in clean text turns. We interrupt, pause, look away, smile, listen, and react. Wan Streamer is built around that product need: a real-time AI avatar that feels present instead of a chatbot placed behind a video renderer.
Voice bots miss the visual layer
A voice-only agent can answer, but it cannot show gaze, expression, timing, or motion. Wan Streamer is designed around synchronized audio and video presence.
Slow video agents break immersion
Traditional cascaded pipelines wait for separate speech recognition, language, speech synthesis, animation, and rendering modules before the user sees a response.
Role play needs a real partner
Language practice, coaching, emotional check-ins, and creative role play work better when the AI can hear, see, speak, and respond with natural motion.
Supported by the Wan-Streamer direction
From cascaded avatars to native streaming interaction
Public Wan-Streamer research describes an end-to-end interactive foundation model for language, audio, and video as both input and output. Instead of relying on separate speech, language, animation, and video-generation modules, the model direction learns perception, reasoning, response timing, and cross-modal synchronization in one streaming system.
Wan Streamer turns that research signal into a focused product landing page: waitlist first, video companion experience next, then pricing, login, credits, storage, and privacy controls after the core beta flow is ready.
Real-time full-duplex conversation
Wan-Streamer introduces a native-streaming interaction model where perception and response are not forced into strict turns. Wan Streamer turns that direction into a product waitlist.
Synchronized video and expression
The product concept centers on AI avatars that answer with voice, facial movement, gaze, and visible listening behavior instead of detached text or audio alone.
One interaction state
The research direction avoids stitching together separate VAD, ASR, LLM, TTS, animation, and video modules, reducing the latency and drift that make avatars feel mechanical.
Real-time AI avatar use cases
Where Wan Streamer can be useful
The early product should focus on high-frequency emotional and conversational needs before expanding into business-facing digital humans. That keeps the first version clear: talk to a visible AI, test the interaction, save the best moments, and share only when you choose.
Late-night AI companion
Open a Wan Streamer live video conversation when you want to talk without waiting for a friend to be available.
Language practice
Practice role play, interview drills, pronunciation, and everyday dialogue with a visible partner.
Interactive digital human
Prototype Wan Streamer-style customer-facing AI hosts, onboarding guides, video concierges, and livestream assistants.
Shareable AI moments
Turn natural conversations into 15-60 second highlights for TikTok, Reels, Xiaohongshu, or Douyin.
Become one of the first Wan Streamer users
Join the waitlist for early access updates, beta character slots, product notes, and Wan-Streamer API availability tracking. The model is not publicly available as a finished product yet, so the first step is building a real audience before wiring payments and credits.
Private Wan Streamer waitlist and early access invites
Character presets for companionship, coaching, practice, and entertainment
Live camera conversation with synchronized voice and expression
Short-clip export for social sharing and creator workflows
Pricing, login, usage credits, and privacy controls after the beta flow is ready
FAQ
Key questions about Wan Streamer, Wan-Streamer, early access, privacy, and the current API status.
What is Wan Streamer?
Wan Streamer is a prelaunch real-time AI video companion and digital human concept for users who want face-to-face AI conversation. The page tracks the Wan-Streamer research direction while preparing a product experience around video chat, voice, expression, and sharing.
Is Wan Streamer the official Alibaba Wan-Streamer product?
No. wanstreamer.org is an independent product page. It references public Wan-Streamer research from the Wan Team / Alibaba Group and will clearly separate research status from product availability.
Can I use the Wan-Streamer API today?
A public production Wan-Streamer API has not been confirmed. The current site is positioned as a waitlist and research-to-product tracker, not as an active API console.
Why is this different from a normal AI avatar?
Most avatar products are cascaded systems: speech recognition, language model, text-to-speech, animation, and rendering are separate. Wan-Streamer points toward native streaming audio-visual interaction in one model, which can reduce latency and improve synchronization.
Will Wan Streamer support Chinese and English?
The landing page is English-first for SEO, but the product direction should support Chinese and English conversations because real-time video companionship, language practice, and creator sharing are cross-market use cases.
How will privacy work?
Camera and microphone features require clear consent, visible controls, and a privacy-first beta. Saved clips should be opt-in, and private conversations should not become public sharing content without user action.
Launch-stage promise: no fake API access claims, no fake model availability, and no hidden pricing until the beta workflow is real.