Text to episode
Text in, episode out. You write a short premise in plain language; the generator writes the scenes, renders each one with a consistent cast and set, and joins them into a single 9:16 video you can watch, publish or add to a series.
What kind of text works
The box takes a premise, not a prompt. You do not need to describe camera angles, lighting or lens choices; the drafter handles the visual grammar and the renderer handles the look. What helps is a person, a place and a problem. "A paramedic on his last shift" is a person; "the empty public pool at dawn" is a place; "finds his own name on the next call sheet" is a problem.
Length of the text matters less than specificity. Two sharp sentences beat a paragraph of adjectives, because the drafter can only build scenes out of concrete details.
How text becomes dialogue
Each scene is written with one spoken line in quotes. The video model performs that line with natural room audio, so the dialogue you read on the review screen is the dialogue you hear in the take. If a line reads wrong, change it there; the rendered performance follows the text.
Voice styles are descriptions, not recordings. A character can be tagged with a style such as low and unhurried or bright and teasing, and that description is added to the take so the delivery leans that way. It steers the model; it does not reproduce a specific voice.
How text becomes pictures
The words in a scene line are only half of what the renderer sees. The other half is a set of reference images: the cast members in the scene, their outfits if any were assigned, the prop if there is one and the set. Those images are what make the pictures consistent from take to take, which plain text alone cannot do.
The preset roster covers the common ground: sixteen characters across ages and looks, twelve sets from a late-night radio studio to a subway car, twelve outfits and twelve props. Anything the text needs that is not there, you can generate in the library with a short description.
From text to a whole series
A single premise makes a single episode, but the structure underneath is a series. Attach the next episode to the same series and the drafter reads the earlier synopses and the last cliffhanger, so the second episode continues the thread instead of restarting it. Over a few episodes, a sentence becomes a season without you writing a document to track it.
Series can be made public. The public page lists the episodes in order with their synopses and plays each published one, which is a reasonable way to share a serial without uploading it anywhere else.
What text cannot change after confirmation
Once an episode is confirmed, its scene lines are fixed; the takes render from them as written. If you want different words, draft the episode again or recut it from the finished takes. This is deliberate: a render is a purchase, and the review screen exists so the purchase is informed.
What you can always change is the episode title, the series title and logline, and whether the result is public.