Vivix.AI announces two interactive video models, A1 and W1, in private beta

Vivix.AI, a small startup, has announced two interactive video models: A1 for full-body characters you can interrupt, and W1 for a live world you shape as it runs, now in a private API beta.

Paylaş
Vivix.AI announces two interactive video models, A1 and W1, in private beta

Vivix.AI, a young startup that describes itself as building real-time interactive video models and the infrastructure to run them, has announced two models in a three-post series on its X account. The company's umbrella message was blunt: "Content is alive." According to Vivix, real-time interactive experiences need two things, characters that come alive and worlds you can step into, so it trained one model for each.

The announcements are the company's own; there is no independent testing or benchmarking, and everything below reflects Vivix's own descriptions and demos. The account opened in September 2025 and remains small.

A1: full-body characters you can interrupt

Vivix describes A1 as a full-duplex, full-body model. Full-duplex (two-way) means the character keeps listening while it talks and moves, so you can jump in mid-sentence or redirect it mid-walk and, the company says, it turns around and follows your lead. Full-body means the character has arms and legs.

  • Cross a room, dance, and pick things up while, Vivix says, objects keep their exact and natural form.
  • Look puzzled when you change your mind, or get hurt when you turn away.
  • "Characters, not just talking heads," in the company's words.

W1: a world that never stops

Vivix-W1 is described as a streaming-native multimodal model, a world model that runs non-stop with no "start over." You shape it as it runs.

  • Talk to it, tap or click the screen, or drop in a picture, and what you do changes what happens next.
  • Everything stays live: characters you can interrupt and worlds you can mess with.

The W1 post included a demo video of a toy-like village scene, with a cursor stepping in as it played. Access is limited for now: Vivix says both models sit behind a private API beta at platform.vivix.ai, not open to the public. The announcements were posted on the company's account, @VivixLabs_HQ; the A1 post links to its model page at vivix.ai, and the W1 post to vivix.ai/vivix-w1.

Devamını oku

Liquid AI تطلق أكبر نماذج LFM2.5 المفتوحة لتشغيل الوكلاء على الأجهزة

Liquid AI تطلق أكبر نماذج LFM2.5 المفتوحة لتشغيل الوكلاء على الأجهزة

أطلقت Liquid AI نموذج LFM2.5-24B-A2B كنموذج مفتوح، ويضم 24 مليار معامل إجمالا لكنه يشغل نحو 2.5 مليار منها في كل خطوة. وبحسب النتائج التي نشرتها الشركة، يولد النموذج المدرب للوكلاء الذين يستخدمون الأدوات 184 رمزا في الثانية على Apple M5 Max.

GlobalFeed Editor tarafından