Most startups now know they need short video ads, and AI video generators have made each clip cheap to produce. Yet many first attempts still look like a slideshow of unrelated scenes: a nice product shot, a random life>
Why a shot list matters more with AI
When you film with a camera, the physical shoot forces you to plan. With AI, you can type a prompt and get a clip in a minute, so it is tempting to skip planning entirely. The result is clips that each look fine but do not add up to a message. A shot list is a short document that says, for every few seconds of the video, what the viewer sees, what they read or hear, and why that moment is there. It keeps the whole ad pointed at one goal.
Step 1: Define one viewer and one promise
Write a single sentence: "For [viewer], [product] helps [result]." For example: "For first-time founders, our invoicing app gets clients to pay on time." Every shot you plan should support that promise. If a shot does not, cut it, even if it looks impressive.
Step 2: Pick a length and count the shots
Short social ads usually run 10 to 30 seconds. As a rule of thumb, plan one shot for every two to four seconds. A 15-second ad therefore needs about five shots, and a 30-second ad about eight to ten. Fewer, clearer shots almost always beat many fast cuts when the product is new to the viewer.
Step 3: Write the shot list as a table
For each shot, fill in four columns:
- Time: for example 0–3 s.
- Visual: what is on screen, written as a concrete scene ("close-up of a phone showing a paid invoice notification").
- Text or voice: the caption or line for that moment, kept under about eight words.
- Purpose: hook, problem, product, proof or call to action.
A typical order is hook, problem, product in action, a detail or proof point, and a closing call to action. Writing the purpose column forces you to notice when two shots do the same job.
Step 4: Keep the visual >
AI clips generated one by one can drift in lighting, colour and even in how the product looks. Before generating, decide on a few fixed >
Step 5: Generate shot by shot, then assemble
You can do this manually with a general video model, generating each row of the table and cutting the clips together in an editor. Some newer tools automate the planning step as well. With Nereo, for example, you describe the video you want and its agent writes the shot list, generates each shot and edits them into one video, which is useful when a team does not have an editor. Whichever route you take, review the plan before the clips: fixing a weak hook in the shot list takes seconds, while fixing it after generation means starting over.
Step 6: Check before you publish
- Does the first second give a reason to keep watching?
- Is the product shown accurately in every shot, with no warped logos or text?
- Can the captions be read on a phone with the sound off?
- Does every claim in the ad match what the product really does?
- Have you followed the platform's rules on labelling AI-generated content?
Step 7: Test one change at a time
Once the first version is live, make variations that change only one row of the shot list, such as a new hook or a different closing offer. Because the plan is written down, it is easy to see what changed and which version performed better. Over a few rounds, the shot list becomes a reusable template for future campaigns.
Conclusion
AI has removed most of the cost of producing video, but not the need for a clear idea. A one-page shot list, written before any generation, is the cheapest way for a small team to get ads that tell a story instead of showing random clips. Start with one viewer, one promise and five shots, and build from there.
