Build a usable shot list, film product B-roll on a phone in about 20 minutes, and learn which script beats need which type of shot for a natural cut.
An AI avatar talking straight to camera for thirty seconds is the fastest way to make a UGC ad feel like a UGC ad shouldn't — flat, static, and obviously generated. B-roll is what fixes that: the product being picked up, opened, applied or used, intercut with the talking-head shots, is what sells the illusion that a real person filmed this with the product actually in hand. This tutorial covers building a shot list before you film anything, capturing usable B-roll on a phone in about twenty minutes, understanding the intercut rhythm that makes footage feel natural rather than randomly inserted, and matching specific shot types to specific script beats so every cut is doing a job instead of just filling time.
Takes 5 minutes
Write down eight to ten specific shots before you pick up a phone: the product on a table, the package being opened, the product being picked up, a close-up of it in use, a close-up of a key detail like a label or texture, the product next to something that shows scale, a reaction-adjacent shot like hands setting it down, and a final packshot. Filming without a list produces beautiful footage that doesn't match any beat in your script.
Write the shot list against your script, not in the abstract. Read through the demo beat specifically and note the exact moment the script needs visual proof — 'watch what happens when I twist this open' needs a matching twist-and-open shot, not a generic product-on-table shot, or the cut will feel disconnected from what's being said even though both pieces are technically about the same product.
Keep the list to eight or ten shots rather than trying to capture everything you can imagine. A focused list you actually finish beats an ambitious list you abandon halfway through, and for a 30-second video you rarely need more than six to eight distinct B-roll shots in the final edit anyway — extra footage mostly sits unused in a folder.
Pro tip:Write the shot list on your phone's notes app so you can check items off while filming. It's easy to forget the detail shot once you're a few products deep into a filming session.
Takes 5 minutes
Film near a window during daylight hours with the light coming from the side or slightly in front of the product, never directly behind it. Avoid mixing daylight with indoor lamps in the same shot, since the color temperature clash is one of the fastest ways to make B-roll look amateurish rather than authentically casual.
A simple test: if you can read the label clearly without squinting and the product doesn't have a harsh shadow cutting across it, the lighting is good enough. You don't need a ring light or a lightbox for UGC-style B-roll — in fact, overly even, shadowless lighting can look more like a studio ad than the intentionally rougher aesthetic that performs well in this format.
Pick one spot near a window and reuse it for every shot in the session rather than moving between multiple lighting setups. Consistent lighting across all eight to ten shots makes the intercut with the avatar footage feel like one continuous scene instead of a patchwork of clips shot in different conditions.
Takes 15 minutes
Hold the phone in your hand rather than mounting it on a tripod, and keep the camera moving slightly — a slow push in, a small pan, or a hand entering and leaving frame. Static, tripod-locked shots read as studio footage; the small, natural camera motion of a handheld shot is what makes B-roll read as 'someone actually filmed this with their product,' not stock footage.
Get closer than feels natural at first. Most people frame product shots too wide out of habit from photography, but B-roll that performs well in UGC videos fills most of the frame with the product and the hands interacting with it — you should be able to see texture, the exact motion of opening a cap, or liquid moving inside a bottle. Wide, distant shots read as packshots, not UGC.
Shoot each shot for eight to twelve seconds even if you only need two or three seconds in the final edit. Extra length gives you room to trim to the exact beat you need and cut out the start and end of the shot where your hand is settling into position, which is almost always the weakest part of any handheld clip.
Pro tip:Film vertical (9:16) or square, not landscape, even if you plan to also make a 16:9 version later. Cropping vertical footage down for other ratios works; stretching landscape footage up to fill a vertical frame never looks right.
Takes 5 minutes
The hook and CTA beats work best with the avatar on camera making eye contact, since that's where trust and action-taking happen. The demo beat is where B-roll should dominate, showing exactly what's being claimed rather than the avatar describing it. The problem-agitation beat can go either way depending on whether a relatable product shot exists — an empty bottle, a messy shelf — or whether the avatar's face carries more emotional weight there.
A rough rhythm that performs consistently well across most product categories is roughly 60% avatar talking-head footage to 40% product B-roll across a 30-second video, with B-roll concentrated in the demo beat and scattered lightly elsewhere rather than clustered at the start or end. Front-loading all your B-roll in the first five seconds, for instance, delays the moment a viewer sees a human face they can trust.
When a script line makes a specific, checkable claim — a texture, a speed, a before-and-after — that's almost always the strongest candidate for a B-roll cut, because showing beats telling for anything visually verifiable. When a script line is emotional or aspirational rather than descriptive, keep the avatar on screen instead, since B-roll can't carry emotional delivery the way a face can.
Takes 5 minutes
Time each B-roll cut to land on a natural action beat in the footage — the moment a cap pops off, the moment liquid starts pouring, the moment a hand lifts the product — rather than cutting at a mechanically even interval like every three seconds. Cuts that land on action feel motivated; cuts on a metronome feel like an editing template.
Scrub through your raw footage and mark the single most satisfying moment in each clip before you start assembling the edit — usually a specific frame where something visibly changes. Building the cut list around these moments first, then filling in talking-head footage around them, produces a noticeably more natural rhythm than assembling shots in the order you filmed them.
Klip Kanvas's editor times the default intercut automatically to land B-roll on demo-beat action when you upload footage against a generated script, which is a reasonable starting point for most videos. Still review the auto-placement once, since it can't always tell which specific frame in your raw clip is the satisfying one — you may want to nudge a cut half a second earlier or later to land exactly on the action.
Pro tip:If a cut feels awkward on playback, the fix is almost always timing, not the shot itself. Nudge the cut point half a second in either direction before reshooting anything.
Takes 3 minutes
Watch the finished intercut on your phone with the sound off before publishing anything. Confirm the B-roll is recognizable at a glance, doesn't get covered by captions or UI overlays, and that the cut points don't create a jarring jump in lighting or framing between shots.
Most viewers decide whether to keep watching within the first few seconds on mute, so this pass matters more than a full-volume review on a desktop monitor. If a B-roll shot only makes sense with the audio explaining it, either the shot needs to be more visually self-explanatory or it needs an on-screen caption reinforcing what it shows.
Good B-roll isn't extra polish on top of an AI UGC video — it's the difference between a talking head that feels like an ad and a video that feels like someone actually used the product. A focused eight-to-ten shot list, twenty minutes of handheld phone filming with real motion, and cuts timed to land on action rather than a clock will get you there without needing any equipment beyond your phone and a window. Build this shot list once per product and reuse the footage across dozens of future script and hook variants — B-roll ages far more slowly than any individual script does.
No. A phone shot handheld near a window, with attention to close framing and natural motion, consistently outperforms polished tripod footage for UGC-style ads. The slightly imperfect handheld feel is part of what sells authenticity.
Plan for eight to ten distinct shots filmed at eight to twelve seconds each, even though a 30-second video typically only uses six to eight of them in the final cut. The extra shots give you flexibility during editing.
Yes — the engine can pull the images and clips it scraped from your product page automatically. New handheld footage generally outperforms static product photos, but scraped images are a reasonable starting point if you have no other footage yet.
Framing shots too wide. Product B-roll that performs well in UGC ads fills most of the frame with the product and the hands interacting with it — pull the camera closer than feels natural at first.
Create your first AI UGC video ad in minutes — no filming, no actors, no editing.
Try Klip Kanvas freeFeed ChatGPT real customer language, lock a four-beat UGC structure, force spoken English, and strip the model tells before anyone films or renders the ad.
Use Gemini in Asset Studio, Demand Gen, and Performance Max as a draft engine: feed a real brief, rewrite every line against claims, and keep human UGC as the YouTube-native asset.
Paste a product link and Klip Kanvas writes the script, casts the creator and renders the ad — no filming, no actors, no editing.
