Hook bank
Prompt-helper + your winning-ad transcripts become a hook list node. 8–12 openings per product angle (unboxing, 3am problem shot, 'I was skeptical', price-drop, dupe comparison).
UGC ads at catalog speed: performer-style video without creator logistics
"UGC creator" draws ~40,500 US searches a month and "UGC ads" ~1,300 at difficulty 5 — because UGC-style creative still out-performs studio ads in Meta and TikTok auctions. But the logistics are the tax: 2026 market rate for a single 30–60s deliverable is a median of $175–$212, 80% of quotes land under $500 (Influee / Influencer Marketing Hub 2026 surveys), and all-in cost lands at roughly 1.5–2.5× the base once you add paid rights, whitelisting, revisions and briefing time. Meanwhile a winning ad burns out in weeks, so a serious DTC account needs 10–30 variants per month per SKU. Berrys produces UGC-style vertical video — performer delivery, your real product locked in frame — from one saved workflow, so creative refresh becomes a batch job instead of a creator pipeline.
Reviewed October 2026 · Prepared by Berrys · Technical specifications & Canvas node workflows
Meta & TikTok UGC-style video specs (checked 22 Oct 2026)
| Parameter / Slot | Marketplace & Production Requirement |
|---|---|
| Format | 9:16 vertical. Export 1080×1920 (Meta accepts up to 1440×2560; TikTok minimum 540×960, 720×1280+ preferred) |
| Meta safe zone | Keep text/logo/key action out of top 14%, bottom 35%, 6% each side — one shared rule for Reels, Stories and vertical feed since March 2026 |
| TikTok safe zone | On the 720×1280 In-Feed template: 160px top, 440px bottom, 80px sides, plus a 120×720px right action rail |
| Length | 15–30s wins completion; TikTok ad format policy is 5–60s (auction In-Feed accepts longer uploads; 21–34s recommended; TopView 9–15s) |
| File | MP4/MOV, H.264 + AAC. Meta ≤4 GB video / ≤30 MB stills; TikTok ≤500 MB, ≥516 kbps (TopView ≥2,500 kbps) |
| Hook window | Product on screen and claim spoken within the first 1–3 seconds — before the swipe decision |
| Spark Ads | Require an organic post from a real TikTok account. An AI persona cannot be whitelisted — run AI creative as standard non-Spark, or book a real creator for Spark |
| Economics | Real UGC: $175–$212 median per video, ×1.5–2.5 all-in with rights. Canvas output is owned on generation — no rights negotiation, no ghosting, no briefing calls |
The creative-refresh pipeline
Performer selection
Pick one persona per brand voice; the face, wardrobe and delivery style lock across the whole batch so your account doesn't look like five different people.
Product lock
Real packshot feeds the SKU node — the bottle/garment/device held on camera is your actual packaging geometry and typography, never a re-drawn lookalike.
Script + lip-sync
Script node writes per-hook 15–25s reads; the video model (Kling 4.0, Seedance 2.5, Veo 3.1) renders delivery, then lip-sync aligns to your VO or TTS track.
Captions & overlay
Text nodes burn platform-safe captions inside the 14/35/6 zone; price/offer badges update from one variable — change the offer once, re-export every variant.
Variant fan
List node fans hook × body × CTA into a matrix (e.g. 4 hooks × 2 bodies × 2 CTAs = 16 ads) with per-placement crops. Ship the matrix, harvest the winner, iterate the winner's hook, kill the rest.
Where generic AI breaks UGC ads
Mid-shot product substitution
Video models drift objects: the bottle in hand at second 1 is a slightly different bottle at second 12 — label text re-flows, pump shape changes. In a UGC ad the product IS the claim, so anything that doesn't lock geometry to a real packshot is unusable. Berrys conditions every render on your SKU reference.
Hands and touch are the trust layer
UGC works because a person touches the product: grip, squeeze, fabric pinch, lid twist. Those are exactly where generative video shows artifacts — warped fingers, objects passing through palms. Hand-heavy beats get built from real reference frames, or the hook is rewritten to show the product on-surface instead of in-hand.
Label the machine, don't be the machine
TikTok requires AI-generated-content labels on realistic synthetic media, and Meta requires disclosure of AI-manipulated creative. Unlabeled AI personas risk enforcement; over-polished AI aesthetics just read as ads and eat your thumb-stop rate. The workflow exports the compliant version: labeled, native-feeling handheld framing, imperfect rooms.
How Berrys produces the set
Brand Memory keeps voices stable
Persona, tone, palette and product truth persist across every campaign so your ad account reads as one brand — the failure mode of single-prompt UGC tools is that every video feels like a different company hired a different creator.
Multi-model video canvas
Kling 4.0, Seedance 2.5, Google Veo 3.1 and Omni Flash under one subscription — you benchmark delivery styles per market without four logins and four bills.
Node graphs, not prompts
The refresh pipeline is saved once. Next month's creative batch is 'swap hooks, run graph' — which is the actual answer to 2–4-week ad fatigue.
Production workflows in Berrys Canvas
16-Variant ad fatigue killer
1. File Node: Load verified master packshot and winning hook bank.
2. Brand System Node: Enforce 14/35/6 safe-zone rules for TikTok In-Feed and Meta Reels.
3. Video Node (Seedance 2.5 / Kling 4.0): Render 20-second vertical performer delivery with product held in-frame.
4. List Node: Fan out 4 hooks × 2 body angles × 2 CTAs = 16 ad deliverables in one run.
Category dupe-comparison hook
1. File Node: Connect your product packshot alongside neutral generic baseline.
2. Brand System Node: Lock brand typography and price badge overlay variables.
3. Video Node: Generate 15-second split-screen hook comparing price-to-value without altering pack geometry.
Seasonal room reskin without reshooting
1. File Node: Connect locked product SKU reference.
2. Brand System Node: Swap room environment preset (e.g. bathroom counter to festive holiday shelf).
3. Image Node (Nano Banana 2.1): Batch render 40 seasonal creative variants while keeping product pixels identical.
Berrys Plugins · Model Context Protocol (MCP)
Create in chat. Keep it on your canvas.
Already working in an AI agent? Connect Berrys to Claude, Cursor, ChatGPT, Codex, or Hermes. Your agent plans the ugc-style ad creative suite and runs the shots on your Berrys canvas, with saved Brand Memory and credit approval in chat.
Connector URL:
https://app.berrys.io/api/mcpRemote MCPDrop-in Agent Brief (UGC-style ad creative)
Using Berrys, generate a 16-variant UGC ad package for my product reference in ./assets: 1. File node: Load product packshot. 2. Brand System node: Set 9:16 vertical layout. Keep text/captions strictly inside the 14% top, 35% bottom, and 6% side safe zones. 3. Video node (Seedance 2.5): Generate 4 distinct 3-second hook variations with natural handheld framing. 4. Image node (Nano Banana 2.1): Render high-res first-frame thumbnail masters at 1080×1920. Estimate credit cost and request approval before generating.
Compliance & production boundaries
AI UGC is a format, not a testimonial. Everything in this workflow is legitimate as demo, explanation, unboxing-of-your-own-product, and scene-setting; faking a customer review, a skin transformation, or 'I lost 5kg with this' with a synthetic person is an FTC endorsement violation and platform policy violation — and it converts worse once people notice. Berrys is built for the format layer: your product truth, delivered at volume.
When NOT to use AI generation for this
- Testimonial or review ads — lived experience must come from a real person
- Health, efficacy or results claims delivered by an AI persona
- Spark Ads whitelisting (needs a genuine organic account)
- Anything where the selling point is verifiable human performance (stain removed on camera, weight held, charge tested)
Frequently asked questions
Is AI UGC legal to run as an ad?
Yes with disclosure. TikTok requires AI-generated-content labels on realistic synthetic media; Meta requires flagging AI-manipulated people/scenes. What you must NOT do is fabricate customer testimonials or imply real lived experience — that's an FTC endorsement problem, not a platform-label problem.
Does it replace my UGC creators?
It replaces the volume layer: hooks you'd never A/B-test, seasonal reskins, long-tail SKUs a creator won't brief on for $200. Keep real creators for testimonial ads, Spark whitelisting and anything that needs a genuine lived experience.
Which models look most like handheld UGC?
In our canvas benchmarks, Seedance 2.5 and Kling 4.0 deliver the most natural delivery cadence; Veo 3.1 wins lighting realism in scene-first shots. You switch nodes in the graph without changing the workflow.
How do I keep the product from morphing mid-video?
Condition every render on the real packshot through the SKU node and keep the product in the lower-center safe zone; re-encode at ≥516 kbps so TikTok compression doesn't shred label typography.
Can I run it as a Spark Ad?
Not from an AI persona — Spark requires an organic post from a real account. Run AI creative as standard In-Feed, or commission the post from a real creator and whitelist that.
Related tool comparisons & guides
In-Depth Guides & Research from Berrys Journal
Create on-brand assets without booking a studio
Start free on Berrys Canvas — 50 credits, no card required — or send Berrys Studio a brief and get finished campaign visuals with agency art direction.