Create Videos Using AI: Who Decides What, in 20 Decisions
How to read the ledger
Three kinds of decision, in the order a project meets them. The AI decides: it makes a choice from your script and you can change it. The AI proposes: it fills something in or drafts it, and you confirm. Only you: nothing is filled in for you, or the choice is yours by nature. Other guides sort the same project differently: Create videos with AI by what you have, and Video AI by which models do which job. This one sorts it by who holds the decision.
The AI decides, and you can override (8)
| # | Decision | What the AI does | How you override it |
|---|---|---|---|
| 1 | Where the episodes break | The reading cuts your script into episodes by story: more script gives more episodes, not longer ones | Cut the script yourself before the read. Afterwards the split is fixed; a different split is a new project |
| 2 | Who is in the cast | Anyone who acts, speaks or is handed something becomes a character with a role: Lead, Supporting or Villain. A part heard and never seen is marked Off-screen | Add, delete or relabel in step 2. Adding a character the script did not produce needs a paid plan |
| 3 | What each person looks like | Writes an appearance and clothing for every character | Edit the text and regenerate the sheet |
| 4 | Which places and props get pictures | Lists the places and the objects that matter to a shot | Add or delete entries in step 2 |
| 5 | The shot list | Writes each episode as segments of shots, with a size, an angle, the line and the seconds | Edit the text (free), change shot seconds, insert or delete a segment |
| 6 | How each person sounds | Writes a voice description per character; by default the video model performs it | Ask for a free voice match and attach a preset you can hear, or upload a voice |
| 7 | The music | Plans a score for the episode once its first shot exists, picks or writes the pieces and mixes them under the voices | Only the Music switch beside the player. There is no track picker or volume control |
| 8 | Joining the segments | Joins the finished segments into one episode file | You press Composite once every segment has a clip |
The AI proposes, and you confirm (6)
| # | Decision | What the AI proposes | Where you confirm it |
|---|---|---|---|
| 9 | The three looks | Pre-fills the visual, scene and costume styles that you have not touched, tagged as AI’s choice | Step 1. All three must be set before the read starts |
| 10 | The script, if you have only an idea | AI Create writes a draft from your line | You read and edit it in a window before it is read |
| 11 | The summary | After the read, step 1 shows the title, protagonist, era, strategy and the counts of characters, places, props and episodes | You check the cast and the episode count |
| 12 | The cast pictures | Draws a sheet for each character, place and prop | Accept, regenerate, AI edit or upload your own. You can continue without making all of them |
| 13 | The next episodes | Let AI continue drafts one to three episodes from where the story stands. AI story continuer | You read and edit the draft, then add it |
| 14 | The cover | Pre-selects a genre mood and your leads, and paints three poster options | You choose one |
Only you (6)
| # | Decision | Why it is yours | Where |
|---|---|---|---|
| 15 | What the story is | An idea, or a script, and how many people and places. Fewer, steadier characters hold up better | The first tab you use |
| 16 | Type: Short Drama, TV Series or Film | It sets how the script is cut, the length cap of an episode (240 s, 300 s, none) and the acting grammar | Step 1. Fixed when the script is read |
| 17 | The shape: 16:9 or 9:16 | The shape of every shot | Step 1 |
| 18 | Dialogue language and the Target Region | The language the cast speaks, and where the names, faces and places are set | Step 1. Fixed when the script is read |
| 19 | Quality and spend, segment by segment | Each segment is priced by seconds and tier. The cheapest tier starts at 10 credits a second | The model button on every segment card |
| 20 | Which takes to keep | You watch each clip, regenerate a failed one, or fix it with Edit this clip: trim, change the visuals, redub | The episode editor. Each segment keeps a version history |
Decisions 16 and 18 are the two that cannot be taken back once the script is read, so make them on purpose. Changing the region, dialogue language or type after the script is read says what to do if you chose wrong.
The ones that lock early
| Decision | Fixed when | To change it |
|---|---|---|
| Type (16) | The script is read | A new project with the same script |
| Dialogue language and Target Region (18) | The script is read | Edit descriptions and lines, or a new project |
| Visual style | You continue past step 1 | A new project |
| Scene style and costume style | The script is read; they are written into the descriptions | Edit the descriptions in step 2 and regenerate the sheets |
| Cast, props, shot text, tier, takes | Never | Any time |
That is why the first four rows get your attention first and the last row can wait. Why scene and costume style are chosen before the read explains the third and fourth.
The order to check them in
The cheapest place to find a wrong decision is before the expensive step. Four checkpoints cost little or nothing:
- The summary (11). Is it the right protagonist, era and number of episodes? The read is priced by length, 4 credits per thousand characters on the default.
- The cast pictures (12). Do the faces look like the people you meant? The first 15 setup images on an account are free.
- The storyboard text (5). Read the shots and fix the lines. Editing the text is free, and episode 1 of the first drama has a free storyboard.
- The preview. One shot of about 5 seconds with your own cast, on the cheapest tier, once per account. A new account also starts with 50 credits.
Only after those does video start, at the price on each card. The longer checklist is Checklist before you generate videos, and Where credits go in one episode adds up a whole episode.
FAQ
How do I create videos using AI?
Start with a script, an idea, product photos or a clip. The AI reads it into a cast, places and shots, you check the cast and the text, and the video is made segment by segment and joined.
What does the AI decide for me?
The episode breaks, the cast list and roles, how people look, the places and props, the shot list, the voice descriptions, the music and the joining. You can change most of them.
What can only I decide?
The story, the Type, the shape, the dialogue language and region, the quality and spend per segment, and which takes to keep.
Which decisions cannot be changed later?
Type, dialogue language and region are fixed when the script is read. The visual style is fixed once you continue past step 1.
Can I change the music?
There is no track picker or volume control. A Music switch beside the player turns the score off in playback.
Do I have to approve the AI’s choices?
For the looks, the script draft, the summary, the pictures, the next episodes and the cover, yes: they are proposals. Nothing that costs credits starts without you pressing a button.
Can I check anything before paying for video?
Yes: the summary, the cast pictures, the storyboard text and a short preview, all free or covered by the free allowances.
How long does one video take to make?
The reading takes minutes and each segment runs on the server while you do other things. Closing the page does not stop a generation.
Make the four early decisions, then check them in order
Type, language, region and look are fixed when the script is read, so they are the ones to settle first.
Updated November 2, 2026