B-roll is where most AI UGC ads are won or lost. The avatar sells the story; the footage proves the product exists. This hub covers uploads, scraped images, stock clips, the right mix, and the quality floor below which nothing looks credible.
01
Uploading Footage
1.Can I upload my own product footage and photos?
Yes, and you should. Upload clips and stills into the project's asset panel and the engine intercuts them between avatar scenes automatically, matching cut points to the script beats. Your own footage is almost always higher-converting than anything generated or licensed, because it shows the exact product a buyer will receive. A practical minimum is six to eight assets per ad: two hero angles, two in-use shots, one packaging or unboxing moment, and one detail macro. Anything beyond about fifteen assets per 30-second ad tends to go unused.
#Uploading Footage
2.What file formats and sizes are accepted?
Standard web and camera formats: MP4 and MOV for video, JPG, PNG and WebP for stills. Keep individual video uploads under a few hundred megabytes and trim long clips before uploading rather than after — a 4-minute source file wastes upload time when the edit will only ever use 3 seconds of it. HEIC files straight off an iPhone should be converted to JPG first; they are the single most common upload rejection we see. Vertical source footage is preferable for 9:16 output, but horizontal clips can be cropped in the editor.
#Uploading Footage#Limits
3.How much of my own footage do I actually need?
For a first test batch, three to five usable clips is enough. The engine can carry a 30-second ad on avatar scenes plus scraped product images alone, but the ads that beat control almost always contain at least one shot of the product being handled by a real hand in a real room. If you have nothing, film three shots on a phone in ten minutes: product picked up off a table, product being used, product held to camera at arm's length. That trio covers roughly 80% of the B-roll needs of a standard UGC ad.
#Uploading Footage#Demo Shots
4.Does uploading footage cost credits?
No. Uploads, storage and re-arranging assets in the editor are free. Credits are consumed when a video renders, based on length and resolution, so swapping one B-roll clip for another and re-rendering the affected scene costs the same whether the clip came from your upload, from your product page, or from the stock library. This is deliberate: we do not want the credit model to push you toward the cheaper-looking creative option. Full re-renders of a finished video do consume credits again, so batch your B-roll changes before rendering.
#Uploading Footage#Stock B-roll
5.Is phone footage good enough, or do I need a camera?
Phone footage is not just acceptable, it is preferable. UGC ads live or die on looking unproduced, and a modern phone shot handheld in natural light reads more authentic than a gimbal-stabilised mirrorless shot with a shallow depth of field. Three rules matter more than the camera: shoot with the light source in front of the product, not behind it; hold the shot for at least four seconds so the editor has trim room; and shoot vertically at 1080p or higher. Deliberate slight handheld wobble tends to outperform tripod-locked footage in feed placements.
#Uploading Footage#Image Quality
6.How should I organise footage for batch generation?
Name files so the system and your team can sort them without opening anything: product-angle-context, for example serum-dropper-bathroom-01. Group per product, not per campaign, and tag each clip as hook, demo, result or packaging. When you generate 20 ads at once, that tagging is what stops the engine repeating the same packaging shot in the demo slot of every variant. Ten minutes of naming discipline before a batch saves an hour of QA after it, and it is the difference between 20 distinct ads and 20 near-duplicates.
#Uploading Footage#Limits
7.Do I need releases for people appearing in my uploaded footage?
Yes, and this sits with you rather than with us. Anyone identifiable in footage you upload — a friend's hands are usually fine, a face is not — needs a signed release covering paid advertising use in your target territories, and that release should be broad enough to survive a campaign running in markets you have not launched yet. Employees are not automatically covered by their employment contract in every jurisdiction. We supply the platform and the asset tooling; the advertiser is responsible for the rights to everything they upload.
#Uploading Footage#Limits
02
Scraped Product Images
8.What does Klip Kanvas pull from my product URL?
When you paste a product URL the scraper collects the product title, description, bullet benefits, price, review text and the gallery images on the page. Those images become the default B-roll pool, and the review language feeds the script generator. On a well-built Shopify or Amazon listing this is usually six to twelve images, which is enough to assemble a full ad without you uploading anything. On a thin landing page with one hero image, expect to supplement it — one image cut repeatedly across a 30-second ad looks like a placeholder.
#Scraped Product Images
9.What if the scraped images are wrong, outdated or belong to a variant?
Delete them from the asset panel before generating. The scraper takes what the page serves, so a listing that still shows last season's packaging or a colourway you discontinued will produce an ad that contradicts your landing page — one of the fastest ways to tank conversion rate even when the ad metrics look fine. Check the asset panel every time you import, and re-scrape after any packaging change. If your PDP uses a JavaScript image carousel, some images may not be captured at all; upload those manually.
#Scraped Product Images#Limits
10.Which product pages scrape best?
Shopify product pages scrape most reliably, followed by standard WooCommerce and BigCommerce templates, because their image galleries sit in predictable markup. Amazon listings usually return the main gallery. Heavily custom single-page sites, pages behind a login, and geo-gated storefronts are the common failures. If a scrape returns fewer than three images, treat that as a signal to upload manually rather than fighting the importer. You can always paste the product data as plain text and attach your own assets — the script quality does not depend on the scrape succeeding.
#Scraped Product Images
11.Are lifestyle images better than white-background packshots?
For UGC ads, yes — decisively. White-background packshots read as catalogue photography and break the illusion that a real person is talking about a product they own. Use them sparingly, ideally only once and late in the ad when you are showing exactly what arrives. Lifestyle images with hands, kitchen counters, bathroom shelves and imperfect lighting keep the ad inside the creator-content frame. A useful ratio is one packshot to every three lifestyle assets, and never open on a packshot.
#Scraped Product Images#Image Quality
03
Stock B-roll
12.What is stock B-roll and when should I use it?
Stock B-roll is licensed generic footage — a hand scrolling a phone, a person waking up tired, a gym interior — used to illustrate the problem half of the script, not the product. Use it for the setup beats only: the pain point, the context, the emotional before. It is genuinely useful when your product is intangible (a service, an app, a supplement whose effect is invisible). Never use it for the product itself. A viewer who sees a generic capsule bottle instead of yours will not believe the review, and the ad's hold rate usually collapses in the 5–10 second window.
#Stock B-roll
13.Does stock footage hurt ad performance?
It can, and the effect is measurable. On accounts we work with, ads built predominantly from stock clips typically show hook rates in the same band as owned-footage ads but hold rates 3–6 points lower, because the drop happens after the viewer starts looking for the product. Our working rule: stock may occupy the first 3 seconds and no more than about 20% of total runtime. If you find yourself needing stock for more than that, the honest conclusion is that you need to film something, not that you need better stock.
#Stock B-roll#Avatar vs B-roll Mix
14.Can I use stock or competitor footage of a product that isn't mine?
No, and this is a real constraint rather than a policy formality. Stock licences cover the clip, not any trademark or trade dress visible inside it, and running a competitor's product in your ad invites both a platform complaint and a trademark claim. The same applies to screenshots of other brands' packaging, celebrity likenesses, and press logos you have not been cleared to use. We provide licensed music and stock libraries and the tooling to keep your assets organised, but the advertiser owns compliance for everything that appears in the final cut.
#Stock B-roll#Limits
15.Can I reuse the same B-roll across many ads?
Yes, and it is one of the highest-leverage things you can do. Save your clips to the shared asset library and they are available to every project in the workspace, so a single filming session powers dozens of variants. The caution is fatigue: if the same three clips appear in every ad a warm audience sees, you are refreshing the script but not the visual, and the audience registers it as the same ad. Rotate the demo section even when the hook changes, and keep at least three interchangeable versions of your core demo shot.
#Stock B-roll#Avatar vs B-roll Mix
04
Avatar vs B-roll Mix
16.What's the right ratio of avatar scenes to B-roll?
Start at roughly 60% avatar, 40% B-roll for a 30-second ad, then adjust by category. Products whose value is visual — cookware, cosmetics, apparel accessories, cleaning products — do better at 50/50 or even 40/60. Products whose value is a claim or an experience — supplements, software, services — need more avatar time because the face is carrying the credibility. What consistently fails is either extreme: 100% talking head reads like a webcam pitch, and 100% B-roll reads like a brand film and loses the UGC signal entirely.
#Avatar vs B-roll Mix
17.Should the first 3 seconds be avatar or B-roll?
Avatar, in most cases. A face making eye contact and speaking is the strongest thumb-stop available in a feed, and hook rate is the metric you are optimising in that window — target 30%+, and treat anything under about 20% as a failed hook. The exception is when the product itself is the pattern interrupt: a genuinely surprising visual result, a satisfying transformation, an unusual object. In those cases open on 1.5 seconds of B-roll then cut to the avatar. Test both; it is one of the cheapest variants you can generate.
#Avatar vs B-roll Mix#Demo Shots
18.How long should each B-roll cut be?
Between 1.2 and 2.5 seconds for paid social. Shorter than about one second and the viewer cannot parse what they are looking at; longer than three and the pacing sags and hold rate drops. The engine defaults into this band automatically, but if you are hand-editing, count the cuts: a 30-second UGC ad usually has 12–18 total cuts across avatar and B-roll. Slow the cuts down for the demo section — that is the one place a viewer wants to look longer — and speed them up under the benefit list.
#Avatar vs B-roll Mix
19.Can I run an ad with no avatar at all, just footage and voiceover?
Yes. Generate the script and voiceover, then set every scene to B-roll — you get a voiceover-over-footage ad, which is a legitimate and well-tested format, especially for products that photograph better than they explain. It also sidesteps avatar-realism concerns entirely. The trade-off is real: voiceover-only ads generally show lower hook rates on cold traffic because there is no face to arrest the scroll, and they need noticeably stronger footage to carry 30 seconds. Use them as a variant in a test batch, not as your only format.
#Avatar vs B-roll Mix#Limits
05
Image Quality
20.What resolution do my images and clips need to be?
1080 pixels on the short edge is the floor for anything that fills the frame. For 9:16 output that means at least 1080×1920 for full-bleed shots; an image at 800×800 will be visibly soft when scaled up. Stills used as a partial overlay or a corner inset can be smaller. Video should be at least 1080p at 24–30fps; 60fps sources are fine and get conformed. If your only asset is a 600px thumbnail from an old listing, use it as a small inset with a caption over it rather than letting it fill the screen.
#Image Quality
21.Can AI upscale a low-resolution product image?
Partly, and you should know where it stops. Upscaling recovers plausible edge detail and removes some compression blocking, which is enough to rescue a 720px image for a 1.5-second cut. It cannot recover text — small print on packaging, ingredient panels and dosage instructions come back as convincing-looking gibberish, which is both an ad-review risk and a trust problem if a viewer pauses. Never upscale an asset whose job is to make a label readable. Re-shoot that one shot instead; it takes two minutes with a phone.
#Image Quality#Limits
22.My product footage looks washed out next to the avatar scenes. How do I fix it?
This is a colour and exposure mismatch, and it is the most common reason an otherwise good ad feels stitched together. Three fixes, in order of impact: apply the project-level colour match so all B-roll is graded toward the avatar scene's white balance; re-shoot under a single light source rather than mixed daylight and tungsten; and avoid heavy phone-camera HDR, which flattens contrast in a way that reads as cheap. If one clip is the outlier, cut it shorter — a 1.2-second cut hides a lot that a 3-second cut exposes.
#Image Quality
06
Demo Shots
23.Which demo shots matter most?
In order: the product doing the thing it is bought for, the result, then the product in hand. Everything else is decoration. For a skincare serum that is the dropper on skin, the skin after, the bottle held to camera. For a kitchen gadget it is the gadget cutting, the finished plate, the gadget on the counter. Budget two shots for the mechanism and one for the outcome. Ads we see win consistently spend 6–10 seconds of a 30-second runtime on demo footage, concentrated between seconds 8 and 20.
#Demo Shots
24.How do I film demo shots cheaply?
One phone, one window, one afternoon. Set the phone on a stack of books near a window with the light falling across the product, shoot each action three times, and keep the takes long. Total kit cost can be zero; a £15 clamp and a £10 diffuser is the only upgrade worth making. Film in one session for the entire product line rather than per-campaign — a 90-minute shoot that produces 40 clips will supply B-roll for months of ad iterations, which is far cheaper per creative than commissioning footage each time you want a new variant.
#Demo Shots#Uploading Footage
25.Can AI generate a demo shot of my specific product in use?
Not reliably, and we would rather say so than let you find out in an ad review. Generative video can produce a plausible generic object in a plausible hand, but it cannot reproduce your exact packaging, logo placement, label text and product proportions with the fidelity a buyer will compare against your PDP. Small errors — a mirrored logo, an extra finger, wrong cap colour — are common and read as deceptive even when they are accidental. For anything where the viewer must recognise the actual product, film it or use scraped photography.
#Demo Shots#Limits
26.What about before-and-after shots?
Use them carefully and own the compliance. Before-and-afters are among the highest-performing B-roll structures in beauty, fitness, cleaning and home categories, and also the most heavily policed by ad platforms — Meta restricts before-and-after imagery for health, weight and body claims outright. Practical rules: same framing, same lighting, same distance, no filters, no zoom change between the two frames, and never generate or composite a fabricated after-state. If you cannot show a genuine result under identical conditions, use a process shot instead.
#Demo Shots#Limits
07
Limits
27.What can the B-roll system not do?
Four things worth knowing before you plan around them. It cannot fabricate a photorealistic version of your specific product in a new scene. It cannot make a low-resolution or badly lit source clip look good. It cannot clear rights on footage you upload, or detect that a clip contains a trademark it should not. And it cannot judge whether a demo shot is misleading — a technically accurate cut can still overstate a result, and ad review will treat that as your problem. Everything else in this hub is a workflow question; these four are hard limits.
#Limits#Image Quality
Ready to put this into practice?
Create your first AI UGC video ad in minutes — no filming, no actors, no editing.
