How to make a SaaS launch video that actually explains the product

Most launch videos do not fail because the animation was bad. They fail because nobody did the arithmetic first, and the arithmetic is unforgiving.

A viewer reads on-screen copy at roughly 160 words per minute, not the 250 or more they manage on a page. Each line arrives cold, there is no going back, and the eye has to find the text on the frame before it can start reading it. That search costs about 900 milliseconds before the word count matters at all.

Put those two numbers together and a scene has a hard capacity:

Write to those numbers and your video is readable. Ignore them, and you have made a video where the copy is technically present and practically invisible.

Start with one sentence, not a script

Before any scene exists, write this down:

For [who], this helps [outcome] without [the friction they have today].

Everything else is a filter against that sentence. A screen that does not serve it belongs in a different video, or in no video.

The most common failure at this stage is trying to cover the whole product. A launch video is not documentation. It is an argument that one specific thing is now easier, and an argument gets weaker with every extra clause.

Five to ten scenes, and why the floor exists

A 30 second film wants five to ten scenes. That is not a stylistic preference, it is the range where each scene still gets enough time to be read.

Below five, individual scenes run long enough that the eye finishes early and starts waiting. Above ten in 30 seconds you are averaging three seconds a scene, which after the 900ms hold floor leaves room for about five words. Five words is a label, not a message.

A workable 30 second shape:

1. The problem. A situation the viewer recognises, with a visible cost in time or attention.
2. The turn. The moment they open the product or bring their material in.
3. The one action. A single gesture, framed large enough to read.
4. The result. What changed, shown rather than claimed.
5. The ask. One specific invitation.

Five beats. Six seconds each. Thirteen words each. That is the whole budget, and it is smaller than most people expect.

Timing is not taste

Motion has bands that track how the eye accepts a change as caused rather than as a jump:

• Under 80ms, a change is not perceived as motion at all. It reads as a different state. Any easing curve you applied to it was discarded.
• An element arrives in 250 to 600ms.
• The subject of the frame gets 400 to 900ms.
• Only a camera move earns 800 to 2000ms.

Exits run at about two thirds of the entrance. A thing arriving is information and earns its time; a thing leaving has already been read, and holding it on screen while it fades makes the frame feel slow for nothing.

Items in a list land 40 to 120ms apart. Past that a stagger stops reading as one gesture and becomes a queue: the last item arrives most of a second after the first, and the viewer has finished with the frame before the frame has finished with itself.

Prepare screenshots like evidence, not decoration

Clean them before importing. Remove real customer data. Zoom into the state you want read. Pick shots that show a progression rather than a tour.

For each one, answer four questions:

• What should the viewer look at?
• What happened immediately before?
• What happens immediately after?
• What sentence would you say out loud over it?

If you cannot answer the fourth in one line, the screenshot is too busy for a short video. Three useful screenshots beat fifteen near-identical ones, every time.

Where the frame is not yours

If the film is going anywhere vertical, the platform draws its own interface on top of your video. Roughly the bottom 22% and the right 10% belong to captions, handles and an action rail.

A call to action placed there renders perfectly and is invisible to the audience. This is the single most common and most expensive mistake in vertical product video, because nothing in your editor warns you about it.

Plan for version two

First versions almost always change. The message moves, a screen gets replaced, the ask needs to be blunter.

Scene-by-scene structure is what makes those edits cheap. You change one scene instead of rebuilding a timeline, and the other four keep their timing. That is the practical argument for a storyboard over a keyframe editor for this kind of work.

---

Scenivia is built on exactly these numbers: it plans five to ten scenes against a 15, 30 or 60 second budget, checks the copy in each one against its reading time, and keeps the safe area off-limits automatically. Join the waitlist.