Most corporate video scripts start as a Word document written in paragraphs. The narration, the CEO’s lines, the text for the lower thirds and a note saying “drone shot of the building here” all run together. That holds up while one person reads it.
Then the shoot comes and the presenter reads “drone shot of the building here” off the teleprompter. In the edit, someone has to guess which sentences were meant to be on screen, and the motion designer types the CEO’s job title from memory.
The two-column script, also called an AV script (audio and video), gives each kind of information its own place. What the viewer sees goes on the left and what the viewer hears goes on the right.
In short
- Put what the viewer sees in the left column and everything the viewer hears in the right one, one row per moment.
- Number the rows, tag each entry in capitals and keep directions out of the spoken text.
- Write lower thirds and numbers in the script exactly as they should appear, and ask the client to confirm those lines.
- At 140 words per minute, a 60-second video holds about 140 spoken words, so time your presenter and cut before the shoot.
What goes in each column
The left column is the picture. It describes the shot (who’s on camera, which B-roll, which product close-up) and holds every piece of text that will appear on screen: lower thirds, full-screen cards, animated numbers, the end logo.
The right column is the sound. Every word someone says out loud goes here, whether it’s voiceover, an interview answer or a presenter talking to camera. Music and sound effect cues go here too.
Each row is one moment of the video. Read across a row and you know what’s on screen while that line plays. Read down the right column on its own and you have the full spoken text.
That’s what the presenter will read on the teleprompter and what you’ll time, and later it’s what you check the final cut against.
The template
Copy the two tables below into Google Docs, Word or anything that handles tables. The first one is the header and goes at the top of page one.
| Field | Fill in |
|---|---|
| Client | |
| Video code and title | |
| Script version and date | |
| Target length | |
| Approved on the client side by |
| # | Video (what’s seen) | Audio (what’s heard) |
|---|---|---|
| 1 | ||
| 2 | ||
| 3 |
Number the rows and keep the numbers when lines are added or cut (5a and 5b are fine). When the client writes “change row 6”, everyone opens the same place, even on a printout.
Fill in the approval line before anyone books a crew. A corporate script often passes through several people at the client, and you want to know whose yes is the final one.
Conventions that keep the script usable
Write one row per spoken line, or per sentence when a line runs long. A row with a whole paragraph of narration forces the video column to describe four shots in one cell, and nobody can tell which picture goes with which sentence.
Start every entry in the video column with a tag in capitals. B-ROLL is supporting footage, LOWER THIRD is a name and title, CARD is full-screen text, GRAPHIC covers animated numbers and diagrams, LOGO is the end slate. The editor scans for B-ROLL, and the motion designer can pull every on-screen text out of the script in a couple of minutes.
In the audio column, put the speaker in capitals followed by a colon: VO for voiceover, a first name for interviews and presenters. Music and sound effects get their own tags (MUSIC, SFX) and a short note on what happens, like “MUSIC: up full for five seconds, then under.”
Keep directions out of the spoken text. A “(pause)” in the middle of a line ends up on the teleprompter, and it inflates the word count as well.
If the presenter needs a direction, put it in the video column of the same row. Our guide on how to read a teleprompter covers the presenter’s side of this.
Write numbers in the audio column the way they’ll be said. “Forty-two thousand square meters” is easier to read aloud than “42,000 m²”, and spelling it out keeps the word count honest. In the video column, write the same number exactly as it should look on screen.
On-screen text goes in exactly as it should appear
A misspelled name in a lower third can survive three rounds of review, because that text gets typed by hand late in the edit, often from an old email. Put the final text in the script and the motion designer copies it instead of retyping it.
For each lower third, write the name the way the person spells it and the job title the client approved (titles change, and the one online may be out of date). Mark where the line should break with a slash.
For numbers, settle the format before the edit: thousands separator, decimals, currency, units. Dates follow whatever format the brand uses. Put it all in quotes, so whoever reads the cell knows that’s the exact text to type.
Then ask the client to check those lines in particular. “Please confirm the names, titles and numbers in the video column” gets a faster and more careful answer than “please review the script”.
Example: a 75-second corporate video
A made-up logistics company opening a new distribution center, with one voiceover, two people on camera and a music open with no speech.
| Field | Fill in |
|---|---|
| Client | Fennick Logistics |
| Video code and title | FNK_014, New distribution center |
| Script version and date | v2, October 21, 2026 |
| Target length | 1:15 |
| Approved on the client side by | Marta Ribeiro |
| # | Video (what’s seen) | Audio (what’s heard) |
|---|---|---|
| 1 | Drone, wide: the new building at sunrise, trucks lined up at the docks. LOGO: Fennick, bottom right. | MUSIC: opening theme, up full for five seconds, then under. |
| 2 | B-ROLL: forklift moving pallets between racks. Slow push-in on the conveyor. | VO: Every day, eleven thousand orders leave this building. Two years ago, it was four thousand. |
| 3 | Marta at her desk, medium shot. LOWER THIRD: “Marta Ribeiro / CEO, Fennick Logistics” | MARTA: For two years we were turning customers away. The old warehouse simply couldn’t keep up, so we built one that could. |
| 4 | B-ROLL: wide interior, all fourteen loading docks. GRAPHIC: “42,000 m²” counts up over the shot. | VO: Forty-two thousand square meters, fourteen loading docks and a sorting line that runs around the clock. |
| 5 | B-ROLL: night shift, boxes leaving on the belt, a truck pulling out at dawn. | VO: An order placed before ten at night is packed, sorted and on the road by six the next morning. |
| 6 | Daniel walking the floor with a tablet. LOWER THIRD: “Daniel Okafor / Director of Operations” | DANIEL: Every package is scanned in under a second. If a label is wrong or a box is damaged, we catch it here, on the line, before it gets anywhere near a truck. |
| 7 | B-ROLL: morning briefing with the team. Close-ups of hands scanning boxes. | VO: Two hundred and thirty people work here across three shifts, and most of them live within twenty minutes of the gate. |
| 8 | Marta, same setup as row 3. No lower third, she’s already been introduced. | MARTA: Our customers want their orders tomorrow. Our people want to get home on time. This building lets us do both. |
| 9 | CARD: “Fennick Logistics / Since 1998” over the drone shot from row 1. | VO: Fennick Logistics. MUSIC: up, then out. |
The audio column has about 145 spoken words, a little over a minute of speech. The music open and the end card bring it to roughly 1:15.
How many words for your target length
Count the words in the audio column only, leaving out speaker tags and music cues. Divide by the speaking pace, then add the moments with no speech, like a music open or the end card.
Pace depends on the person. A presenter reading from a teleprompter moves at a different speed than a CEO answering a question, and the same person speeds up when nervous.
The table below uses 140 words per minute, the baseline RecReview uses for its duration estimate. Time your own presenter reading one row out loud and adjust. Pace by type of video, and how to time a script properly, are in how many words fit in a minute of video.
| Target length | Spoken words at 140 per minute |
|---|---|
| 30 seconds | 70 |
| 60 seconds | 140 |
| 90 seconds | 210 |
| 2 minutes | 280 |
| 3 minutes | 420 |
| 5 minutes | 700 |
If the script comes in long, cut words before the shoot. Trimming a line on paper takes a minute. In the edit it means cutting into a take that may have no clean place to cut.
For interviews, write the answer you expect, note in the video column that it’s a guide, and count it anyway. Once the length is settled, it’s also the number you plan the shoot day around.
How this works in RecReview
In RecReview the script lives on the video’s task. You can upload a file (PDF, Word, PowerPoint, TXT or RTF, up to 50 MB) or link a document from Google Drive or OneDrive.
If your team writes in Google Docs, save this template as a Doc, connect Google Drive and set it as the workspace’s script template. On a task, Create from template copies it into the folder you chose and names the file from a pattern, such as the video code plus the title.
It also fills any placeholders you typed in the header, like {{client}}, {{project}}, {{code}}, {{title}}, {{date}} and {{duration}} (the target length set on the task). The new Doc is attached to the task as its script.
Calculate duration reads the script, keeps only the spoken text and counts it at 140 words per minute. The result becomes the video’s estimated length on the task.
When the video goes into a production plan, its shoot block takes that length and multiplies it by the pace you pick for the day (Realistic, the default, is 3×). The estimate counts speech only, so you add a five-second music open yourself.
The same cleanup feeds the teleprompter. It drops lower third and card text, camera directions, scene headings, sound effects and the speaker names in front of each line, so the presenter sees only what they have to say. You can still edit the text before recording if something slipped through.
When the cut comes back, AI Review can compare it with the script. It listens for lines that were dropped, added or changed and notices sections that were skipped or moved.
The lower third text isn’t part of that comparison, since the script goes through the same cleanup as the teleprompter, so names and titles on screen still need the client’s confirmed list.
Each finding comes with the timecode and what was actually seen or heard there, and the editor decides what to change. The whole script is checked on the first cut. Once the client has commented on a version, their notes become the reference instead.
