AI Video Generator

What goes in

There are four ways in, one per tab on the home page. They differ in what you hand over and in what the first step costs; after that they share one path.

The clip is read, not reshot: the video model never receives your upload. It is cut into shots and described, and a new video is generated from that description with your cast. What you can feed in has every format and limit; Four ways to start a project helps you pick a tab.

What happens in between

Whichever tab you used, the project goes through the same five stages. Each one produces something you can read before the next one is paid for.

1. The read. The text is read for characters, scenes and props, split into episodes, and given a visual style. It takes minutes and you can leave the page. What comes back is a summary (title, protagonist, era, how the story was adapted), an episode list, and a description for every character, location and prop.
2. The cast. Each description becomes a picture: characters at 9:16, locations and props at 16:9. The first 15 images on an account are free. These pictures are the reference every later shot is made from.
3. The storyboard. Each episode becomes segments, and each segment becomes three to five numbered shots with their seconds, the people and props named, and the lines with a voice. It takes a few minutes. Editing it costs nothing.
4. The segments. A segment is one generation. You press the button with the price on it, the credits are held, and the clip arrives in typically ten to twenty minutes.
5. The join. The segments are put together into one file per episode, at 1 credit a second.

The order puts the cheap, readable stages first on purpose: a wrong name or a missing second is a free edit in stage 3 and a paid regeneration in stage 4. What is inside a project shows how the pieces nest.

What comes out

The badge rules are in AI video without a watermark. For the numbers behind each row, see What comes out.

A real minute, costed

Take a one-minute episode on the default tier. Video is 60 seconds at 10 credits a second, and joining is 60 seconds at 1 credit a second. Together that is 660 credits, about $4.11 at the lowest credit price, before the read, and before any retake. The read of a short script is a few credits more, and the storyboard of your first episode is free.

The bigger a project, the more of the bill is video. The setup, the storyboard and all the editing together are a small share of a series; the lever that moves the total is how many seconds you render and at which tier. The AI video cost calculator itemises an eight-episode series.

For a sense of scale, the sample projects on this site are single episodes made by this same path, and they run from about a minute and three quarters to three and a half minutes, with casts of four to eleven characters. The Last Envelope has six characters and two sets and runs 140 seconds in 16:9; The Violin on Mercer St is the vertical one, at 148 seconds in 9:16.

Where the time goes

Almost all of it is the model. The stages above take minutes, not seconds, and the video stage is the longest. A reading usually takes a few minutes; a setup image about a minute and a half; a segment typically ten to twenty minutes. An internal measurement in August 2026 averaged around nine and a half minutes for a segment, so treat the card’s own estimate as the live number.

How many generations run at once is a property of your plan: 3 on a free account, 5 on the first paid plan, more above that. Start more and the extra ones wait, credits held, and start by themselves. You can close the tab; the work continues on the server. Why a generation takes time.

What it will not take in or give back

• A link instead of text. Pages are not fetched. Paste the text.
• A PDF. Refused when you pick it. Export to text or Word first.
• One button that animates a single still. Photos become an ad, or the face, room or object that a drama’s shots follow. Image to video AI has the two routes.
• Readable words drawn into the picture. Video models do not render text reliably, so what must be read goes on an end card or is spoken.
• A recognisable public figure or a franchise character. The check refuses it at generation.
• Free video without limit. Video is charged from the first second; the free part is the read, the first images, one storyboard and one short preview. What is free.

What decides how it looks

Four things, none of them a magic word. The tier sets resolution and the longest segment. The visual style you pick in step 1 sets the look, and it cannot be changed once you continue. The setup images set who everyone is. And the storyboard text sets what happens, which is why most odd shots are fixed by editing a sentence, not by paying for a higher tier. No magic words and Describe the cause, not the result cover the writing.

Start on the cheap tier. Both takes stay in the segment’s history, so shooting the keepers again on a higher tier costs one more generation and nothing else. Draft at the lowest tier, finish at the highest.

FAQ

Updated November 2, 2026