How to make an AI sales roleplay video
Brief for AI Create: Four two-minute training scenes for a team selling route-planning software to delivery fleets: the buyer says the price is double, the timing is wrong, a competitor is cheaper, and ‘just send me something’; each scene shows a weak reply, then a strong one; one buyer, one rep, one glass meeting room.
Two things people mean, and the one this is
| Search intent | What it needs | SceneMixer |
|---|---|---|
| An AI buyer to practise against | Live conversation, branching, scoring | No: a video is the same every time you press play |
| A video of a sales conversation to show a team | A scene, two people, a script | Yes |
| A set of scenes for a course | The same cast across many scenes | Yes: a TV Series project with one buyer and one rep |
Use the video for what video is good at: showing, once and identically, what a good reply sounds like, so a team can discuss it. AI corporate training videos as short dramas.
Start with the objection table
Write the table before the script; each row becomes one scene.
| Objection | The buyer says | Weak reply | Strong reply |
|---|---|---|---|
| Price | Your price is double what we pay now. | Offers a discount at once | Asks what a missed delivery costs them today |
| Timing | Come back next quarter. | ‘Sure, I'll email you.’ | Asks what has to be true by then, and who decides |
| Competitor | The other tool is cheaper. | Criticises the competitor | Asks which part of the other tool they use most |
| Brush-off | Just send me something. | Sends a brochure | Asks one question that earns a ten-minute call |
Keep the product and company invented. Brands and logos are not drawn reliably, and well-known brands trigger the IP check; describe the space instead. The IP check.
One scene, shot by shot
The price objection, in the storyboard's own format. One 15-second segment, four shots:
[Shot 1] 3s: Medium shot, a glass meeting room. @Dana, a fleet manager, arms folded. @Dana (flat): Your price is double what we pay now. (voice: Dana)
[Shot 2] 4s: Over-the-shoulder on @Ravi, the rep, leaning in. @Ravi (warm, even): It is higher. What does a missed delivery cost you today? (voice: Ravi)
[Shot 3] 3s: Close-up on @Dana. @Dana (pausing): Nothing. We just apologise. (voice: Dana)
[Shot 4] 5s: Medium shot, @Ravi opens a notebook. @Ravi (calm): Then let's price the apology. (voice: Ravi)
Each line is short enough for its shot. Write the same scene twice, once with the weak reply and once with the strong, and they share the room and the people, so only the behaviour changes. Writing dialogue in the storyboard.
A word budget per scene
English runs about three words a second, and a segment's seconds are all the speech has to fit in, with nobody doing anything else.
| Scene length | Seconds | Spoken words, at most |
|---|---|---|
| One segment | 15 | about 45 |
| A long take | 30 | about 90 |
| A one-minute scene | 60 | about 180 |
| A two-minute scene | 120 | about 360 |
Real calls ramble. Cut every line that is not the point. Shot seconds and timing.
Same two people, many scenes
Make the four objections episodes of one TV Series project (up to 300 seconds each; two minutes is plenty). The buyer and the rep are drawn once and the glass meeting room is drawn once, so Dana looks like Dana in every episode. Why a series stays consistent · The same room in every episode.
Versions for other teams, and delivery
- Languages. Make the episode again with another dialogue language and region. A second language version · Fifteen dialogue languages.
- Subtitles. Burned in, or an SRT file for a learning platform, on accounts that have made a purchase. Subtitles.
- Review. Colleagues can review through a share link, which opens without an account and always carries the AI badge. Using SceneMixer as a team.
Make it in SceneMixer, start to finish
- Script: paste your own on Upload Script, or write the brief above in AI Create and let it draft the episodes.
- Step 1: Type TV Series; Scene Style Urban Opulence; Costume Style Professional Uniforms, Smart Casual; Visual Style Digital HD; Target Region; dialogue language; 16:9 Landscape. Then read the script.
- Step 2: Draw Dana, Ravi and the glass meeting room; they are the whole cast and the whole set of every episode. The first 15 setup images on an account are free.
- Storyboard: generate episode 1 first (free on your first drama) and read it before you render anything. One place per segment.
- Video: render on the default tier at 10 credits a second, fix single shots instead of whole episodes, then put the episode together and post.
| Episode | Objection | The strong reply's one move |
|---|---|---|
| 1 | Price | Ask what the problem costs them now |
| 2 | Timing | Ask what has to be true by then |
| 3 | Competitor | Ask what they use most |
| 4 | Brush-off | One question that earns the call |
FAQ
Can AI make a sales roleplay video?
Yes, as a scripted scene: a buyer and a rep, one objection per episode, weak and strong replies in the same room.
Is it interactive?
No. A video plays the same every time. It shows a conversation; it does not take part in one.
Can I use our real product and logo?
Describe the space instead. Logos are not drawn reliably and well-known brands trigger the IP check.
Can I make it in other languages?
Yes. Make the episode again with another dialogue language and region.
How long should a scene be?
Two minutes is plenty; at about three words a second a two-minute scene holds about 360 spoken words.
Can I add subtitles for a course?
Yes: burned in or as an SRT file, on accounts that have made a purchase.
Can I use recordings of real calls?
Upload only material you have the right to use, with the consent of every real person whose voice appears. A recording of a real person is that person's. How voices work.
What does it cost?
Video is 10 credits a second on the default tier plus 1 credit a second to join; a two-minute scene is 120 seconds of each, before retakes.
Updated November 2, 2026