XGEN-JING · Community showcase

A world from
your point of view.

Explore first-person video and audio. Compose a scene, map its movement, and prepare your next experience with JING.

Free showcase & prompt editor

Official prerecorded demo · XGEN Labs · This page does not generate video.

Compose an experience

Your scene. Your next move.

Each scene contains a prompt and a sequence of movement slices. Export a ready-to-run cases file for the official inference setup.

Scene builder

Output video name

Seed

Reference image paths · optional, up to fiveOne path per line on your inference machine. Refer to images as <Picture 1>, <Picture 2>, etc. Images are not uploaded here.

Inference case

Updated as you edit. Fixed resolution: 960 × 544.

1scenes

3slices

56frames

How to generate your video

Install the official XGEN-JING inference code on a supported GPU server. The published setup is validated on six H100 GPUs.

Place your reference images at the paths above, save your cases file, then run:

bash demo.sh cases=/path/to/cases.json

The editor derives frame count using the default model layout: 17 × total slices + 5. Longer sequences require more compute. This is batch inference; the released causal model is still forthcoming.

A world fromyour point of view.

Your scene. Your next move.

Scene builder

A world from
your point of view.