Skip to main content

W 1.1

November 6, 2026

Model Features

  • Auto mode. W1 Auto selects a generation speed automatically, making it easier to get started without choosing a speed for each request.
  • Video references. Use a reference video to guide motion, camera movement or visual style, alongside your prompt, images and audio.
  • More natural faces. Improved skin texture and facial highlights reduce the overly glossy look.
More natural facesSkin texture, expressions and highlights as the lighting changes.
Video reference: cat cooks fried riceA new subject follows the cooking actions from a reference video.

API Updates

  • Auto: set model to vivix-w1-auto when submitting a video generation job.
  • Video upload: upload reference videos with POST /v1/files/videos.
  • Video input: pass the upload response’s data.url in jobs[].ref_videos as an array of { "url": "<UPLOADED_VIDEO_URL>" } objects. Describe what to borrow from each video in your prompt. Video references cannot be combined with first_frame_image or last_frame_image.
See the video reference request example and field details →

W 1.0

October 9, 2026

Model Features

W 1.0 is officially released, bringing low-cost, fast video generation to creators and developers. Create more variations and iterate on your ideas with three generation speeds: Normal, Fast and Ultra Fast.
SneakersClose-up tracking follows each step across a sunlit court.
Potato chip adProduct details and dynamic movement in a short commercial.

API Updates

Select a model option using the model field:Submit a job with POST /v1/videos/jobs, then query its status with GET /v1/videos/jobs/{job_id} to retrieve the generated video.Read the World Video Generation guide →