A restaurant menu of 20 dishes as 6-second 9:16 clips: $15 to $17

Twenty dishes as 6-second vertical clips cost $15.00 on Sume when each clip starts from your own photo. If Nano Banana 2.1 makes the still first at 1K ($0.10), each dish is $0.85 and the menu is $17.00.

The clip is a 6-second Gemini Omni Flash 1.1 job at 720p, 9:16, which is $0.75 and comes with generated sound.

Two ways to start

Both routes end in the same clip. They differ in what the diner sees on the plate.

Use your own photo for dishes you serve

A generated dish is a picture of a plausible dish, not the one the kitchen plates. Portions, garnish and colour will differ, and diners will compare. For items on your menu, shoot a phone photo and use it as the Omni first frame. Keep generated stills for mood images: a table setting, a cocktail pour, a dining-room scene.

Google's Business Profile help says photos should have no significant alterations and should represent reality, per the help page. Apply the same test to a clip you plan to put there.

Running the batch

Twenty dishes are twenty independent jobs. Name each Idempotency-Key after the dish, such as dish-margherita-v1, so a retry is traceable and a re-roll is v2. Poll status for each, then fetch results.

Plan one re-roll for about a quarter of the dishes. With five re-rolls at $0.75, the own-photo menu lands at $20.00 rather than $15.00.

Practical limits

Steam, sizzle and pour motion are what Omni does well in food clips, but the model can also add or drop garnish mid-clip. Watch each clip once at full size before it goes on a menu page.

Sume does not write menus or prices into the picture reliably, and does not post the clips. Keep the price list in your menu system.

Rates are the Sume card on 2026-10-07. Check pricing_skus and GET /v1/images/models for live rates.