How to Make a VSL: A Step-by-Step Guide

A Video Sales Letter isn't a fancy explainer video. It's a structured, story-driven pitch engineered to hold attention and move a stranger to buy. Get the structure right and the production almost takes care of itself. Here's the process I use on real projects, start to finish.

Step 1: Get the offer and audience clear first

Before a single word of script, you need brutal clarity on two things: exactly who you're talking to, and exactly what they get. A VSL written for "people who want to be healthier" converts nothing. A VSL written for "men over 45 who've been told their blood pressure is creeping up and are scared of medication" converts. The tighter the avatar, the sharper every line that follows.

Step 2: Write the script around a proven structure

Almost every high-converting VSL follows the same skeleton: pre-hook, hook, lead, problem, agitation, mechanism, proof, offer, and close. You don't invent a new structure, you fill a proven one with your specifics. We break each of those beats down in the anatomy of a high-converting VSL, and there's a fill-in-the-blanks version in our VSL script template.

The script is 80% of the result. Nail this and a mediocre edit still sells. Skip it and the slickest production won't save you.

Step 3: Obsess over the first 15 seconds

Most VSLs are lost in the opening. If your hook doesn't create a curiosity gap or name a raw problem in the first few lines, nobody sees the rest. Write five completely different openings, not one. You'll test them later, see these VSL hook examples for patterns that reliably stop the scroll.

Step 4: Record the voiceover

You have three options: record it yourself, hire a spokesperson, or use a high-quality AI voice. Direct-response VO isn't radio-smooth, it's conversational, urgent, and a little imperfect. Whatever route you choose, the pacing of the read sets the pacing of the whole VSL.

Step 5: Produce the visuals

This is where the method matters. Traditionally, every visual means footage you have to shoot or license. With AI production, each line of the script becomes a purpose-built shot, a mechanism animation, an ingredient close-up, a before/after, a location you could never afford to film, generated and revised the same day. If you already have great footage or a strong on-camera presence, traditional editing may be the better path. Most projects use both.

The visual on screen should never fight the voiceover for attention. It should make the exact idea being spoken impossible to misunderstand.

Step 6: Add captions, on-screen text, music and pacing

Word-level captions keep sound-off viewers watching. On-screen text reinforces the three or four ideas that matter most. Music sets emotional tempo. And ruthless pacing, cutting every half-second that drags, is what separates a VSL people finish from one they abandon.

Step 7: Test more than one version

This is the step most people skip, and it's the one that actually decides your winner. A single VSL is a guess. Run different hooks, leads and closes against each other and let the data pick. Because AI makes variations cheap, testing stops being a luxury, see how to A/B test a VSL.

How long does making a VSL take?

Traditionally, weeks per version. With a directed AI pipeline, a first finished VSL in days, with variations following quickly. Speed only matters because it lets you test while the market is still hot.

Frequently asked questions

Do I need a script to make a VSL?

A script is the foundation of the entire VSL, it's where 80% of the result is decided. If you don't have one, a direct-response copywriter should be your first hire. From there, the voiceover and every visual are built around the words.

Can I make a VSL without filming anything?

Yes. With AI production, every scene is generated from the script, no camera, location or spokesperson required. This is ideal when you need to show things that are hard or impossible to film, like mechanisms, ingredients or transformations.

How many VSL variations should I test?

Start with several hook and lead variations rather than one polished cut. Most serious offers test 5-10+ versions across hooks, leads and offers, because the winner is found through testing, not guessed in advance.