The Plush Lipstick Influencer Review recipe is scenes you swap, not a new Omni board

/workflows/plush-lipstick-influencer-review is a locked scene list. Recreate it; do not storyboard it again in Kling Omni.

The public recipe is a 9:16 selfie-vlog: Sara reviews the Plush Baby Pink Lipstick in three scenes — greeting, product review with features, promo-code CTA — with consistent voiceover across the cut. That is a scene graph: prompts, references, model, durations. Opening Kling 3 Omni and asking for "six shots of an influencer reviewing lipstick" is a new film that happens to contain the word review.

Three selfie-vlog scenes, two locked assets

The workflow object is specific:

• Aspect: 9:16
• Scenes: three, eight seconds each, twenty-four seconds of finished tape
• Cast / product: Sara (influencer character) and the Plush Baby Pink lipstick — locked reference stills, not "a girl with a tube"
• Chain: chain_from_previous is false on every scene. These are discrete I2V beats, not last-frame continuity
• Model on the recipe: Vidu Q3 Image to Video
• Credit estimate: 186 at preview resolution, 372 at full

Scene one is the greeting still: Sara in the Scandinavian room, warm smile, no product required. Scene two puts the matte black tube with gold Plush embossing in her hand. Scene three brings the tube closer for the code. Voice lines are already written on the recipe. You do not re-board them in Omni syntax.

Each scene is a pair: a still, then that still animated, then a stitch. Read a workflow's credit estimate before you run it: scene count is not the bill. The video model is.

Recreate copies the graph; Omni storyboards a bumper

Recreate a viral template is the agent verb: browse the public catalog, clone scenes, style, and model into your account, then run or remix. Cloning is free. Running is priced like the generation it is. You see the cost before anything dispatches. It is saved but not auto-run — swap the lipstick still, swap the VO, keep the cuts.

Recreate the workflow, not the screenshot is the same rule in conversion form. A screenshot of a UGC review tells you the caption was big. It does not tell you Vidu Q3, the two assets, the three eight-second beats, or 186 / 372.

Kling 3 Omni multi-shot is not Movie Mode is the other confusion. Omni takes one prompt and returns up to six distinct shots inside a single generation, usually a 15-second file. Identities have to survive inside that generate. There is no scene planner, no three-beat VO pass, no combined 24-second export of greeting / review / CTA. You get one bumper with cuts in it.

Useful for a three-beat gag that does not contain your SKU. Not useful for Sara-plus-lipstick with locked refs. Omni restages. Restaging is drift. Drift on a lipstick is a wrong tube.

Swap the lipstick stills. Keep the eight-second cuts.

Once the format works, the workflow is the product. Clone it. Change the SKU still and the three voice lines. Keep the selfie-vlog room, the 8-second beats, the greeting → features → code shape. Do not go back to an Omni storyboard every Monday because last week's shade performed.

If the clip you care about is not in /workflows, you do not have a recipe. Build one: stills first, then I2V, then save your graph. Edit the clone before you run it. The expensive miss is cloning and auto-running on the stock Plush tube when the client SKU was the point.

FAQ

Is Recreate the same as generating "influencer lipstick review" in Omni?

No. Recreate clones the public workflow — three 8-second scenes, Sara plus lipstick refs, Vidu Q3 Image to Video, 186 / 372 when you run it. Omni is one generate with internal cuts, typically ~15 seconds, up to six shots. Different objects. Different files.

Does cloning the Plush template spend the 186 credits?

No. Clone is free. Credits apply when you run it, at the cost shown before dispatch. Swap the product still and the promo line while it is still a draft.

Can I use Omni under the hood and keep the three-scene structure?

Not as this recipe. Omni multi-shot lives inside one short generate. The Plush review is three discrete I2V scenes with locked refs, stitched, not chained off last frames. Even if a scene happened to render on a Kling family model, the object is still the scene list, not Omni storyboard syntax.

Why is a 3-scene recipe 186 / 372 when some longer recipes list less?

Because the estimate is counting the video model, not the scene count. Vidu Q3 Image to Video across three still-plus-motion pairs is a different unit from a cheaper I2V row across fifteen scenes. Read the model line on the workflow page before you approve the run.