Script
Each line names who speaks, and can carry a note such as a smile or a pause.
LeVik Studio / The Engine
We are building a show engine.
We give AVSP a script, a picture and a voice for each person, and it renders the video itself. Every frame is decided by us.
What goes in, what runs, what stays.
A script, a picture and a voice for each person, and a picture of the studio go in at one end. A finished video comes out the other. Inside, the episode moves through stations in a set order, each handing its work to the next, so a better voice tool next year changes only that one station.
The inputs
Portraits
A picture of every person in the show. The face each viewer watches is built from it.
Voice
A voice reference for each person, so every line is spoken in that voice.
Studio picture
A picture of the room. It becomes the set the show plays on.
What runs for every episode
-
Everything goes in
The script, the portraits, the voices, and the studio picture.
-
01
Voice
Speaks every line in that person's voice, and records when each word is said.
-
02
Face
Turns each line's audio into face movement: lips, jaw, brows, blinks, and emotion.
-
03
Body
Posture, gesture, and the listener: nodding, looking at the speaker, reacting.
-
04
Direction
Who sits where, how the light falls, which camera shows whom, and when to cut.
-
05
Render
The frame is drawn by our own pipeline, not generated from a prompt.
-
06
Edit
The shots and the sound are joined into the final video.
-
07
Check
Automatic checks first, such as lip sync. Then a person watches the whole video before anything ships.
-
A finished episode comes out
Picture and sound, joined and checked, ready to watch.
A prompt gives you a video you cannot direct.
Ask a model to make your show and it hands you something you cannot direct or keep consistent, so we render instead.
Directed
Every camera move and cut is chosen and written down. Change one line of dialogue and only that shot re-renders; the rest stands.
Consistent
The host's face and voice stay the same in every episode. That consistency is part of why we render.
Ours
The renders and the archive all belong to the studio.
AI helps inside the pipeline, but the video is ours to direct.
We use AI where it works well: it speaks every script in each person's voice, and it reads a face's movement from sound. Nothing in the frame is left to chance.
-
AI speaks the lines
Each voice is made once and then reused in every episode.
-
We direct the render
Movement comes from the words and the voice, and every frame is rendered under our direction.
First, thirty seconds.
The first thing the engine is making is a thirty-second announcement. See the show it opens.
About the studio: the studio page · Want to know when the show is out? [email protected]