Vivix

What is Vivix

Vivix is an AI creation platform focused on real-time interactive multimodal models. Its core products include:Vivix-A1, the world's first real-time full-duplex model, capable of generating interactive AI characters with full-body movements that can listen, speak, and act;Vivix-W1, a streaming native multimodal model, an interactive narrative engine that can change the storyline in real time through voice, text, and touch, and the real-time video runtime Vivix IVI. Vivix has developed MJD distillation, NVFP4 quantization, and the VMI inference engine, enabling near-30B parameter models to run in real time on consumer-grade GPUs with interaction latency below 1 second. Vivix-A1 has opened its API.

Vivix's main features

• Full-body character generation: Vivix-A1 can generate in real time full-bodyAI digital humans。
• Full-duplex real-time interaction: Users can interrupt by voice at any time, and the character's movements, expressions, and responses change immediately, with no need to interrupt and regenerate.
• Synchronized audio-visual output: Voice, expressions, gestures, and ambient sound are natively modeled as a unified whole, with no signs of stitching.
• Interactive narrative generation: Vivix-W1 can generate a continuously evolving audio-video story from a single sentence, and users can change the storyline in real time.
• Multimodal input: Supports participation in interaction through multiple methods such as voice, text, images, clicks, and touch.
• Live commerce applications: Digital hosts can walk freely and display products, and operators can adjust the livestream direction in real time by voice.
• Ultra-low latency: First-frame latency is less than 0.6 seconds, interruption response is 300–900 milliseconds, and end-to-end latency is less than 1.7 seconds.
• Real-time operation on consumer-grade hardware: A model with nearly 30B activated parameters, through MJD distillation and NVFP4 quantization, can perform real-time inference on a single consumer-grade GPU at a cost of less than $1 per hour.

How to use Vivix

• Register to obtain an API Key: Visit the Vivix API open platform https://platform.vivix.ai/ , register an account, and create an API Key in the developer console. All requests use https://api.vivix.ai as the base URL and carry Bearer authentication.
• Choose a model: Choose the A series (real-time Avatar interaction or script-driven Avatar video) or the W series according to your needs, such as vivix-a1-stream、vivix-w1-stream , etc.
• Create a session: Via the REST interface POST /v1/realtime-avatar/sessions, configure parameters such as mode, character appearance image, voice, persona instructions, and speech recognition (such as server_vad).
• Establish a real-time connection: After obtaining the returned session_id , control the session through the WebSocket control channel (wss://api.vivix.ai/v1/realtime-avatar/control), and audio-visual output is transmitted via RTC (TRTC) streaming.
• Drive interaction: Write the user's voice or text into the dialogue context. The Avatar generates replies in real time and outputs them in a streaming manner as voice, video, and word-by-word subtitles, with support for interruption midway.
• Real-time intervention: Through the Streaming Control interface, you can switch images/characters during generation, send action instructions, or directly script audio, without interrupting and starting over.
• W1 world gameplay: Drive the plot with open events (Event). Events can be canceled before taking effect. Sessions support hold/pause and resume, and new events can be inserted at any time to change the story direction.

Vivix's core advantages

• True Full-Duplex Interaction: Breaking through the "half-duplex" request-response model, the model can continuously generate audio and video while catching new input at any time and changing content instantly.
• Full-Body Character Mobility: A1 is the first real-time model supporting full-body motion generation; characters can walk, turn around, and pick up objects, not just lip-sync from a fixed camera position.
• Native Multimodal Architecture: Speech, text, and images are modeled in a unified signal space, bypassing the ASR+LLM+TTS cascade; perception directly drives behavior, and audio, visuals, and motion are naturally synchronized.
• Ultra-Low Latency: First-frame latency < 0.6 seconds, interruption response 300-900 milliseconds, end-to-end < 1.7 seconds, interaction latency below 1 second.
• Large Parameters + Consumer-Grade Real-Time Operation: Nearly 30B activated parameters combined with an 8x8x4 high-quality compression ratio, through MJD distillation and training-time NVFP4 quantization, achieving over 10,000 video tokens/s on a single consumer-grade GPU, with inference cost below $1 per hour.
• Natively Interruptible Streaming Architecture: Causal generation + VMI codec-decoupled design; new instructions are appended directly to the decoding pipeline without interrupting and restarting the current generation.

Vivix's Comparison with Similar Competitors

Vivix's Application Scenarios

• Live-Stream E-Commerce: Digital hosts appear full-body, walk around, try on and pick up products; operations staff adjust scripts and product selection via real-time voice without interrupting the livestream.
• Virtual Companions / AI Social: AI girlfriend/boyfriend applications, with motions, expressions, and voice naturally synchronized; users can interrupt conversations at any time and characters respond in real time.
• Interactive Film and Television / Interactive Short Dramas: Input a sentence to generate a story that evolves in real time; the audience participates in plot decisions through voice and clicks, opening a viewing mode where "everyone is a director."
• Game NPCs and Open Worlds: Empowering game characters with autonomous perception and action capabilities; players can naturally interact with NPCs via voice, driving open-world narratives.
• Intelligent Customer Service and Virtual Shopping Guides: Full-body digital employees online 24/7, capable of demonstrating products, answering questions, and responding to customer needs in real time, embeddable in websites/Apps/offline large screens.