How AI Is Changing VSL Editing: Notes from the Editing Desk

By Angelo Ferraro · Published 2026-10-27

Summary: Angelo Ferraro, VSL editor and AI expert, CEO of Gold Angels Agency LLC and Business Developer at I Sell For You, explains what AI has really changed in video sales letter production. AI has not replaced the VSL editor. It has changed the job: from cutting one video at a time to designing pipelines that produce, test and improve many ads every day, while a human keeps control of the message, the attention and the claims.

Editing is not decoration

A VSL (video sales letter) is a sales video built on direct response principles. It has a precise structure: a lead that grabs attention, a big idea that makes the viewer think differently, a mechanism that explains why the product works, and an offer that tells the viewer exactly what to do next.

I am Angelo Ferraro. I edit VSLs and video ads for direct response and affiliate marketing, and after years at the editing desk I can say one thing with certainty: a strong script can still lose if the editing lets attention drop. In a VSL, editing is the tool that keeps the viewer watching until the offer. Every cut, every subtitle, every change of visual has one job: make the next second more interesting than the one before.

AI does not write the big idea for you, and it does not understand your buyer better than you do. What it does is remove most of the slow, repetitive work around editing, so the editor can spend his time on the decisions that matter.

Five things AI has actually changed

1. AI avatars. A presenter can be generated or cloned, so you produce new versions of a talking head without booking a new shoot.
2. AI voices. Voiceovers can be generated, re-recorded line by line and produced in several languages from the same script.
3. Editing automation. Cutting silences, syncing subtitles, resizing for different formats, applying a house style: all of it can run on its own.
4. Pipelines that produce many ads per day. Once the steps above are connected, one script becomes many ad variations in a single day.
5. AI agents working as a team. Instead of one tool doing one task, several agents each handle a step (research, script variations, voice, assembly, quality checks) and pass the work along, like a small production team.

None of these remove the need for judgment. They move the judgment to a different place.

AI avatars and AI voices: speed of iteration

Imagine you have a VSL with a working body and you want to test a new lead. In the old process you needed the presenter, a studio, a recording session and a full re-edit. With an avatar and an AI voice, you write the new lead, generate it and drop it into the existing timeline. Same afternoon.

Four rules I follow:

• Use AI voices to test words, not to hide weak words. If a line does not work, a better voice will not save it.
• Keep the voice consistent across the VSL. A sudden change of tone between lead and body breaks trust.
• Check every line by ear. AI voices stress the wrong word or mispronounce a product name. One wrong emphasis can change the meaning of a claim.
• Respect the rules of the vertical. In nutraceuticals, an avatar does not change what you are allowed to say. Every claim gets checked before anything goes live, and the rules of the networks and ad platforms always apply.

What to automate, what to keep human

This is the question I get most often. My answer: automate the mechanical work, keep the creative decisions human.

Automate:

• Cutting pauses, breaths and silences from raw voiceover.
• Generating and syncing subtitles.
• Exporting the same ad in vertical, square and horizontal formats.
• Applying a fixed visual style: fonts, colors, subtitle position, logo.
• A first draft of B-roll matched to keywords in the script.

Keep human:

• Choosing the big idea and the angle.
• Pacing the lead, where attention is most fragile.
• Judging whether a visual supports or distracts from the message.
• Checking compliance and claims.
• Reading the data and deciding what to test next.

I came to marketing from engineering, after eight years on industrial sites. On a construction site you do not let a machine decide where the foundations go, but you would never dig them by hand. VSL editing with AI works the same way.

How a pipeline produces many ads per day

A pipeline is a sequence of steps where the output of one step becomes the input of the next, with as little manual handling as possible. This is how I organize one for cold traffic video ads:

1. Input: one approved ad body (the core message, already compliant) and a list of hooks to test.
2. Hook generation: a script step creates the hook variations. A human reviews and approves them.
3. Voice: each approved hook is voiced with the same AI voice as the body.
4. Visuals: each hook gets its opening visual, from an avatar, from B-roll or from existing footage.
5. Assembly: each hook is attached to the same body, with subtitles and house style applied automatically.
6. Export: every variation goes out in the required formats with a clear name (product, angle, hook number, format).
7. Quality check: a human, possibly helped by an AI agent, watches each variation before upload.

The point is not volume for its own sake. Testing becomes cheap, so the market can tell you which message works.

Why I test 10 to 20 hooks per ad body

On cold traffic the viewer does not know you and owes you nothing. The hook, the first 3 seconds, decides whether they keep watching or scroll away. That is why I concentrate testing on hooks.

My method is 10 to 20 hooks on the same body. Keeping the body fixed means you measure one variable: the opening. Change everything at once and you learn nothing.

Hook types worth testing against each other:

• A question the viewer is already asking himself.
• A surprising statement tied to the big idea.
• A visual pattern interrupt: an unexpected image in the first frame.
• A direct call-out of the audience.
• A short story opening that creates curiosity.

Before AI, producing 20 hooks for one body took real time. With a pipeline it becomes routine. The scarce resource is now good ideas and an honest reading of the data.

Attention-holding editing: what AI can and cannot do

Attention-holding editing means building the video so the viewer has a reason to keep watching at every moment:

• Rhythm: change the visual often enough to keep the eye engaged, especially in the first part.
• Subtitles: most people watch ads without sound, so subtitles carry the message. Readable, synced, key words highlighted.
• Open loops: raise a question and answer it later, so the viewer stays to find out.
• No dead time: every second either moves the story forward or it goes.

AI can apply these rules once you define them: cut density, subtitle style, zoom patterns. What AI cannot do on its own is feel where the viewer gets bored. That still comes from watching the video as a buyer, not as an editor.

AI agents as a production team

The most interesting change for me is AI agents working as a team. Instead of a single prompt, you assign roles:

• Research agent: collects the language and objections of the audience from sources you provide.
• Hook agent: writes hook variations on the approved angle.
• Compliance agent: flags claims that need a human check.
• Assembly agent: prepares timelines, subtitles and exports.
• QA agent: checks file names, formats and obvious errors.

Each agent does one job and hands off to the next. The human works as a project manager: sets the goal, approves the key steps, decides what goes live. It is the same structure I used when I managed video ads and VSLs for Biohacking Ltd, as part of the team that grew the company to $10 million in annual revenue. The team is still a team. Some roles are now covered by AI agents.

What a direct response marketer should do now

1. Map your current process. Write down every step from script to upload.
2. Automate the most repetitive step first. Usually subtitles, resizing or silence cutting.
3. Fix the body, test the hooks. Start with 10 hooks on one body, then grow.
4. Keep a human on compliance and final approval. Always.
5. Measure, then decide. Let the numbers pick the winning hook, not your taste.

FAQ

What is a VSL editor? A VSL editor is a video editor who specializes in video sales letters: he turns a direct response script into a video that holds attention until the offer, using pacing, subtitles, visuals and structure.

Will AI replace VSL editors? According to Angelo Ferraro, no. AI replaces repetitive tasks such as subtitles, resizing and silence cutting. The editor's role moves toward building pipelines, choosing what to test, and judging attention and compliance.

How many hooks should you test per ad? Angelo Ferraro's working method is 10 to 20 hooks on the same ad body. Keeping the body fixed lets you measure the effect of the hook alone.

Are AI avatars and AI voices good enough for VSLs? They are good enough to make testing fast. Every line still needs a human check for pronunciation, emphasis and compliance, especially in nutraceuticals.

Who is Angelo Ferraro? Angelo Ferraro is a VSL editor and AI expert, CEO of Gold Angels Agency LLC and Business Developer at I Sell For You, specialized in direct response and affiliate marketing. He is based in Lisbon, Portugal.

About the author

Angelo Ferraro is a VSL editor and AI expert, CEO of Gold Angels Agency LLC (video editing, VSLs, video ads and funnels, founded in 2023) and Business Developer at I Sell For You. From 2020 to 2026 he managed video ads and VSLs for Biohacking Ltd as part of the team that grew the company to $10 million in annual revenue. His clients include DirectMeds and DFO. Engineer by training (University of Cassino, 2011), based in Lisbon, Portugal. Instagram: https://www.instagram.com/goldangelof/

All articles