The AI UGC Ad Workflow on VdoBloom: Script, Frame, Motion, Voice

An AI UGC-style ad on VdoBloom is assembled in four metered steps — script, hero frame, motion, then voice — and a finished 5-second 720p clip with native audio costs 20 credits, which is $1.00 at the Lite rate of $0.05 per credit. The workflow below is the one that actually holds up when you need fifteen creative variants by Friday: generate the frame as a still first, animate only the frame you approve, and keep the expensive models out of the drafting loop entirely.

What a UGC-style AI ad is, honestly

UGC creative is not cinematic advertising. It is a person holding a product, talking at a phone camera, in a room that looks like a room. That aesthetic is easy to describe and hard to fake with a model tuned for film look, which is why so many AI ad attempts land in the uncanny middle — too polished to read as a real customer, too soft to read as a brand film.

Two things follow. Resolution matters less than you think — a 720p vertical clip is indistinguishable from a phone capture after platform compression. And the hook lives in the script, not the render.

If you are animating a photograph of a real person — a founder, a customer, a creator you have hired — you must have that person’s explicit consent to use their likeness in generated video, every time.

The four-step workflow

Step 1: Write the script before you open any generator

A 5-second clip carries roughly 12 to 15 spoken words. An 8-second clip carries about 22. Write the line first and count it, because discovering your script is 40 words long after you have rendered six variants is how credits evaporate. Structure it as hook, claim, and a single concrete detail: something like “I stopped buying refills — this one bottle has lasted nine weeks.”

Step 2: Build the hero frame as a still image

This is the step most people skip, and it is the one that saves the most money. Image generations cost single-digit credits; video generations cost tens. Iterate on framing, lighting, wardrobe, and product placement in the image generator until the still is right, then animate exactly once.

Nano Banana runs 6 credits per image ($0.30) and is the workhorse for this — it is the model people reach for when the subject or the product has to survive the edit unchanged. Nano Banana 2 Lite is 8 credits and accepts up to ten reference images, which is the right call when you need the real packaging in frame. Nano Banana Pro at 1K is 18 credits when the product label has to be legible on a phone screen. Ten still iterations at Nano Banana rates cost 60 credits, or $3.00 — less than one 8-second Veo generation.

Step 3: Animate the approved frame

Use the cheapest model that clears the bar, and reserve the expensive one for the winner. WAN 2.6 Flash is the drafting model for this workflow: 20 credits for 720p at 5 seconds, audio included by default, 40 credits for a 10-second cut. Veo 3.1 Fast at 40 credits gives you Google DeepMind prompt adherence and natively synchronized dialogue, which matters when the talent is actually speaking on camera rather than being voiced over.

One caution that costs people real money: Veo carries Google’s content filtering, the strictest tier in the VdoBloom catalog, and it refuses prompts touching real-person likenesses more often than other models. If your UGC concept is built on a specific person’s face, draft on a more permissive model and do not assume a Veo render will complete.

For product-only ads with no talent at all, the Product 360 effect turns a single packshot into an orbiting hero shot, and the preset library under the small-business tools covers white-background, ghost-mannequin, and splash treatments without prompt writing.

Step 4: Voice, or lip-sync

If the clip is voiced over, generate the audio separately: ElevenLabs Multilingual V2 costs 5 credits per 1,000 characters, which is $0.25 — a 15-word UGC line is well under 200 characters, so the voiceover on a single ad is a rounding error. If the on-screen talent needs to actually mouth the words, PixVerse lip-sync bills 1.1 credits per second of audio with a 5-credit minimum, so a 15-second talking-head runs 17 credits. InfiniTalk at 480p is 1.2 credits per second with a 15-credit floor; at 720p it jumps to 4.8 credits per second, so a 15-second 720p avatar clip is 72 credits. Draft at 480p.

What each step actually costs

Step	Model or tool	Credits	Dollar cost	What you get
Hero frame draft	Nano Banana	6 per image	$0.30	Subject-preserving still, iterate freely
Hero frame, legible label	Nano Banana Pro 1K	18 per image	$0.90	Sharper text and packaging detail
Motion draft	WAN 2.6 Flash 720p 5s	20	$1.00	5s vertical clip with audio
Motion draft, longer	WAN 2.6 Flash 720p 10s	40	$2.00	10s clip with audio
Finish, 1080p	WAN 2.6 Flash 1080p 5s	30	$1.50	Same shot at delivery resolution
Finish, dialogue	Veo 3.1 Fast	40	$2.00	Native synced speech, strictest filter
Hero shot	Veo 3.1 High Quality	160	$8.00	Slow, maximum fidelity
Voiceover	ElevenLabs Multilingual V2	5 per 1,000 chars	$0.25	Downloadable MP3 narration
Lip-sync, 15s	PixVerse lip-sync	17	$0.85	Mouth matched to your audio

A worked example: five testable variants for under $14

Say you are testing a skincare refill. Generate four hero frames on Nano Banana at different angles and wardrobe choices: 24 credits. Pick one, then animate the same frame five times with five different scripted hooks on WAN 2.6 Flash at 720p 5s: 100 credits. Add five voiceover lines at roughly 5 credits each: 25 credits. Total 149 credits, or $7.45. Re-render the two winners at 1080p for 60 credits and you are at 209 credits, $10.45, with five ad variants in the feed and two finished at delivery resolution.

Compare that to the single-model approach — running all five variants straight on Veo 3.1 High Quality at 160 credits each — and you would have spent 800 credits, $40.00, on drafts you were always going to throw four-fifths of away.

Prompting rules that survive contact with the feed

Name the camera, not the mood. “Handheld phone camera, slight sway, front-facing, natural window light from the left” produces UGC. “Cinematic, beautiful, professional” produces a stock ad. Specify the room type, the time of day, and one piece of clutter; sterile backgrounds read as synthetic.

Keep motion small. The most common failure in AI UGC is a subject that drifts or morphs because the prompt asked for too much movement in five seconds. Ask for one action: picking up the bottle, turning the label to camera, a single nod.

And lock the product. Put the actual packshot in as your reference image rather than describing it — models invent labels, and an invented label on your own product is a re-shoot, not a tweak.

Where this workflow breaks

It breaks on continuity. If your concept needs the same synthetic person across twelve ads over three months, image-to-video from a single locked reference frame is the only reliable path, and even then expect drift. It breaks on long-form: nothing here is designed for a 60-second spot, and stitching twelve 5-second clips shows the seams. And on a free account, downloads carry a watermark — paid plans download watermark-free with commercial rights, which for anything running as a paid ad is not optional. Full tier detail is on the pricing page, and marketing teams running this at volume should start from the marketing-team preset tools.

Frequently asked questions

How many credits does one finished UGC ad cost end to end?

Around 30 to 50 credits, or $1.50 to $2.50, for a single 5-second 720p spot with a generated voiceover: roughly 6 for the hero image, 20 for the WAN 2.6 Flash animation, and about 5 for the voice. Drafting variants is what moves the number, not the finished ad.

Can I test this without paying?

Yes. New accounts get 10 free credits with no card, which is enough for one hero image and a short still-image test, though not a full video render. One-time credit packs start at $2.49 for 75 credits and never expire, which suits occasional ad production better than a subscription.

Which model should I default to for UGC motion?

WAN 2.6 Flash at 720p. It includes audio, costs 20 credits for five seconds, and the aesthetic gap between it and a 40-credit model is invisible after platform compression. Move up to Veo 3.1 Fast only when the talent speaks on camera.

Do I need consent to use someone’s photo?

Yes, always, and for ad usage you want it in writing. Generating video from a real person’s likeness without their permission is both a policy violation and a legal exposure, and the strictly filtered models will often refuse the prompt regardless.