Most performance teams do not need a Hollywood pipeline. They need a UGC product ad that reads as phone-native: a creator holding the SKU, a kitchen or desk that looks lived-in, a hook in the first second, and a packshot before the loop ends. On Picturesque, that stack is three deliberate steps — still, motion, optional lip-sync — without leaving the platform.
Step 1: Lock the Hero Still (Image Studio)
Generate at 9:16 unless the placement is square. Nano Banana Pro is the default when a human and a product share frame — it holds face geometry and label detail better than budget models on the second and third re-roll.
Reference pattern that works:
- @Image1 — product photo (label readable)
- @Image2 — talent reference (pose and wardrobe)
- @Image3 — environment mood board (not a literal copy)
``` @Image2 holds @Image1 toward camera, soft window light, modern kitchen background inspired by @Image3, handheld phone aesthetic, shallow depth of field, authentic skin texture, no studio backdrop ```
Draft 6–10 stills. Pick one where the product label and hands are both sharp. This still becomes the start frame for video — do not animate until the frame is client-approved.
Budget alternative: SeeDream 4.5 at 5 credits for exploration, then a single Nano Banana Pro pass using the best SeeDream output as a reference.
Step 2: Animate for Vertical (Video Studio)
| Goal | Model | Why |
|---|
| Multi-beat story, hands + product motion | Kling 3.0 | Multi-shot and strong I2V obedience |
|---|---|---|
| Fast cheap motion test | Hailuo 2.3 Fast | Draft vertical clips before spending on flagship |
| Locked still, longer xAI motion | Grok Imagine Video 1.5 | Up to 15s I2V at 1080p |
| Trend dance or gesture on a locked portrait | Kling Motion Control | 10 credits — copy reference motion onto the approved still |
Duration: 5–7 seconds for paid social; use Kling 3.0's flexible grid rather than padding to 10 seconds of dead air.
Open Kling 3.0 · Hailuo 2.3 Fast
Step 3 (Optional): Talking Hook
When the ad needs a spoken line ("I did not expect this texture"), run Talking Avatar on a tight crop of the talent face from the still or first video frame. Keep copy under eight words for sync quality.
Use this for the hook only — let the product motion clip carry the middle and CTA card.
Credit Sanity Check (Typical 9:16 Ad)
| Step | Typical cost |
|---|
| 8× Nano Banana Pro drafts | ~120 credits |
|---|---|
| 1× Kling 3.0 5s I2V | ~60 credits |
| 1× Hailuo 2.3 Fast draft (optional) | lower tier |
| 1× Talking Avatar hook | varies by duration |
QC Before Export
- Label text still readable after video (if not, regenerate still with sharper product ref)
- Fingers not fused on product hold — common failure; tighten prompt to "both hands visible, natural grip"
- First frame matches thumbnail — platforms judge the still before autoplay
Where This Fits on Picturesque
This workflow is the same stack we use for template previews and marketplace creatives: image consistency first, video second, avatar hook last. For batch variants (three hooks, one product), duplicate the workflow in Workflows so each branch keeps the same product reference node.