> ## Documentation Index
> Fetch the complete documentation index at: https://docs.wji.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Script to video
> Walk through the pipeline from script to finished segments.
A program moves through five steps. You can click back to a step you have already completed, but not
skip ahead.
## Creating the program
**New program** asks for a name (up to 80 characters, unique within the folder), the dialogue
language, the aspect ratio, and your choice of presenter, scene, and outfit.
Aspect ratios: 21:9, 16:9, 4:3, 1:1, 3:4, 9:16. The default is 16:9.
## Step 1 — Script
Paste the script, or upload a `.txt`, `.md`, or `.docx` file. Older `.doc` files are not accepted.
## Step 2 — Break into segments
Choose **15 seconds** or **30 seconds** per segment — those are the only two options — and optionally
add an instruction such as "keep it conversational".
**Re-split** generates the segments, each with dialogue, an action, an estimated duration, and a
tone.
From here you can edit any line of dialogue, split a segment in two at the cursor, add a segment
below, and delete segments. Tags can be added, and **Suggest tags** proposes some.
Re-splitting **replaces every existing segment**, including dialogue you have edited by hand. WJI
warns you before it does.
## Step 3 — Voiceover
Tick the segments you want, then:
* **Improve or translate dialogue** with AI
* Generate voiceover for one segment or for the whole selection
* Listen back, switch between takes in the voiceover history
* **Approve** the ones you are happy with
Each segment can override the program's tone and speech rate. Tone offers eight options: serious,
composed, warm, cheerful, energetic, dignified, brisk, and conversational. Speech rate runs from
-50 to 100, where 0 is normal speed.
## Step 4 — Video
Each segment becomes a card. You can:
* Optimize the video prompt, then generate
* Change the model, duration, motion, resolution, and reference images per segment
* **Upscale** a result
* **Approve** it
Duration defaults to the voiceover length plus one second, with a five-second floor. Motion offers
still, natural, active, and animated — all of them keep the presenter blinking naturally.
A segment cannot be given video until its voiceover exists. The status shows **missing voiceover**
until you generate it.
Bulk generation opens a progress dialog that confirms first, then reports each segment as it
finishes. It cannot be closed while it is running.
## Step 5 — Finish
Three tabs: **Segments**, **Generation history**, and **Credit report**.
Export options:
| Export | Produces |
| --------------------------- | ---------------------------------------- |
| **Download all voiceovers** | A ZIP of `001-audio.mp3` and so on |
| **Download all videos** | A ZIP of `001-video.mp4` and so on |
| **Download all assets** | A ZIP with `audio/` and `video/` folders |
| **Download stitched video** | One MP4, joined in your browser |
| **Export report JSON** | The full manifest |
| **Export CSV** | The generation history |
The CSV exports **the currently filtered history**, not everything, with columns for time, elapsed,
type, segment, operator, prompt, and credits.
## Status and cost
Video statuses: approved, generating, missing voiceover, failed, generated, no preview, pending. A
retry counter appears on segments you have regenerated.
There is no automatic retry — press generate again yourself.
**Failed generations are still charged**, and are included in the program's cumulative credit
total. The credit report's success rate counts failures and retries in its denominator, which is
why it can look low on a program that eventually came out well.