Granola: the window bag cannot invent a cluster size

I2V the pouch you shot. Overlay weight. Do not generate a new fill.

A granola listing is a window you already printed: the film, the lockup, the clusters the buyer can see through the cut-out. Cluster size is a product fact. Generating a new fill invents a different SKU. Photograph the pouch. Animate that photograph. Letter the weight after lock. Do not prompt a cousin mix.

The window is the SKU

The bag is not atmosphere. It is a die-cut that shows this bake: cluster size, oat colour, inclusion mix, the fill line you actually pack. Distributors compare the window to the carton. A paragraph that says “artisanal granola pouring in slow motion, golden clusters” is not a spec. It is a request for someone else’s oven and someone else’s cut.

Photograph the pouch large in frame. Include the window and the lockup if the mark must read. Light it so the clusters read as clusters, not as a brown smear. If the still would fail a listing audit — crushed corner, a hand covering the window, a fill you do not ship — stop. Motion will not rebake the batch.

Text-to-video has no plate. It will sample an average breakfast bag: bigger clusters, a shinier film, a flavour word that is not on the pouch. Average is a cousin. A buyer who already has the bag on the counter will notice a different cluster before they notice your logo.

Do not generate a cleaner fill “for hero.” Cluster size is like a durometer. A model asked to “show chunky clusters” will melt the window into a different bake. That is a claim.

Seedance I2V holds the pouch; it does not restock it

Image-to-video keeps frame one as the file you sent. Labels, the window cut, and the clusters in that cut survive because they were already pixels. The prompt is for camera and physics: a slow push on the window, a tilt that does not invent fruit, a hold. Re-describing the bake in the prompt is how the model starts drawing a second fill.

Seedance 2.5 is the family page for that still-to-motion job: 4s–30s, 720p ceiling, 47 credits listed, native audio. Image-to-video is the still-to-motion job on that family; the published URL is the family page. One locked frame into a continuous clip. Optional last frame when the move has a destination — front window to three-quarter gusset — only if both stills are the same SKU.

Thirty seconds in one pass is how you do not stitch four short takes and watch the clusters morph at the cut. 720p is the ceiling. A 30-second miss of the wrong bake is a long video bill.

Native audio is on. Write the mix — bag film, a quiet table, no voice — or you get a kitchen band the pouch never had. Quoted net weight does not become a readable title.

Prompt camera only. “Slow push on the window, clusters static, no new inclusions.” Stop. “Make the clusters chunkier” is a new fill. A last frame of a different flavour is a morph. Two photos of two bakes is two SKUs.

Overlay the weight; never generate the fill

Net weight, serving size, “clusters” as a claim: those strings live on the panel you already print. A video model that letters “12 oz” on the film is fabricating a pack. You cannot edit a prompted glyph without a new generate.

Add a text overlay to a video burns one authored line for the whole clip — top, center, or bottom. Type the weight exactly as the panel states. It does not transcribe. Position is the three-way enum, not a free pixel. Swap the string for the next SKU without rebaking the window.

Do not:

• Prompt “show large clusters through the window” when the still shows small ones.
• Ask the model to “fill the bag fuller.”
• Typeset flavour or weight inside the generate.

If the window in the photograph already shows the wrong bake, reshoot. I2V will faithfully move whatever you feed it. That is the point of locking the pouch first.

Not the ecommerce UGC page, not a bakery crumb

Ecommerce brands already have a storefront page: first three seconds, UGC-style review energy, hook tests. That page is a conversion surface. A granola window is a pack-shot problem. Do not rewrite the ecommerce playbook into a pour. You can run a UGC cut later, with a real person and a real grant. This job is the pouch.

Bakeries is nearby and still not this listing. That page is crumb pull, a case at opening, pastry you sell by the piece. A granola brand ships a bag. Do not copy the crumb playbook onto a window still.

Restaging the pouch in a generated kitchen is how cluster size drifts. Hold the window. Buy camera on the photographed fill. If the model grew almonds you do not roast, you generated a new fill. Throw it away. Shoot the bag you ship.

FAQ

Can I text-to-video “chunky clusters” instead of uploading the pouch?

You can. You will invent a bake. Cluster size is the SKU. Upload the window bag to image-to-video on Seedance 2.5 and describe only the camera.

Does Seedance 2.5 add native audio on this family?

Yes. Native audio is on. Write the mix, including quiet. Quoted weight in the prompt will not typeset a panel you can defend. Overlay the line after lock.

Why overlay weight instead of letting the model letter the film?

Because weight is a claim. Overlay type is editable when the pack changes. In-frame type is a new still every time legal reprints the panel, and a wrong glyph is the frame a buyer locks onto.

Is this the same as an ecommerce UGC ad?

No. Ecommerce brands is hook-and-review conversion. This job is I2V of the photographed pouch, overlay for weight, no new fill. A talking reviewer is a different file and a different grant.