A multimodal line for short dramas going global
Compute, models and the editor share one system. One account calls 100+ models, and rendering sits next to the audience you publish to — no stack to assemble yourself.
Rewrite the script
Break into shots
Lock cast and sets
Render a rough cut
Voice and lip-sync
One team wires the data hall to the canvas, so a fault is not passed between vendors.
Where the work happens: an infinite canvas that wires image, text, audio and video nodes into a flow you can run again.
Major closed and open models sit behind one interface, so routing, billing and unit cost settle in one place.
POST /router/v1/models/invokeSame endpoint — change the model field, not your codePick a model
The request this sends
{
"model": "model-name",
"input": {
"prompt": "Rewrite this passage as the opening dialogue of a 30-second episode, and mark the shot changes."
}
}Owned clusters plus managed nodes, placed around the markets short dramas release into, so video jobs are not waiting for cards.
Five steps in production order. What one step hands off is what the next one consumes — nothing is exported and brought back.
A web novel is rewritten as a script, then broken into shots with camera, mood and lighting notes.
Turnarounds and cinematic environments render in seconds on the canvas, with local repaints for detail.
Stills plus motion instructions come out as HD clips in batches, assembled into an automatic rough cut.
Dialogue is performed, then dubbed into English, Japanese, Indonesian and more, with mouths that keep time.
Character libraries, art presets and past footage stay in your space and are reused across concurrent shows.
All five steps share the same assets. Change a character once, and every later shot follows.
Teams whose output is paced by a slate or an ad calendar feel the difference first.
Short-drama platforms
High-frequency slates that cannot run dry.
Mid to large sellers
Batches of multimodal ads built to be clicked.
Art outsourcing studios
Industrial volumes of assets, with characters and sets that stay true.
Render nodes sit against the markets short dramas are watched in, so producing and delivering a cut never takes the long way. A fourth region is still being built.
North America
The largest overseas audience
Southeast Asia
The fastest-growing market
Japan and Korea
High spend, heavy localisation
Characters, art presets and past footage stay in private space and carry across projects, so nothing is rebuilt from scratch.
One interface and one bill, instead of an account and a reconciliation cycle per vendor.
Compute is placed around the markets you release into, so rendering and delivery both happen near the audience.
Our in-house short-drama work runs on the same line, so problems surface here first and get fixed here first.