How do I make a Thanksgiving takeout menu video for my restaurant?

To make a Thanksgiving takeout menu video, take two photos you already have of the food, animate each for 6 seconds with gemini-omni-flash-1.1 as a first-frame image-to-video job, narrate a 380-character script, lay both over a music bed in Timeline 1.0 and burn the order-by date in with caption cues. On Sume that is $0.75 per 720p clip, $0.02 for the voice, $0.125 for the bed, $0.10 for the render and $0.20 for the captions: $1.945 for a 12-second vertical ad. Rates were read on 2026-10-07.

Thanksgiving 2026 is Thursday, November 26, so takeout orders usually close in the week before it. Because the order-by date and the pickup times are exact facts, they go in captions that you write, never in a prompt.

Start from your own photos

Use real photos of your real food: the carved bird on its board, the table of sides. A first-frame image pins the opening of the clip to the picture, so the food on screen is the food you cook. Pass the photo in frame_images with frame_type: first_frame, and ask for something small and slow in the prompt: steam rising, a hand placing a dish, a gentle push-in.

gemini-omni-flash-1.1 takes 3 to 10 seconds at 360p, 720p, 1080p or 4K, in 16:9 or 9:16, with native synced audio. At 720p it is $0.125 per second on Sume, so 6 seconds is $0.75, which is already a whole cent amount and needs no rounding.

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: tg-takeout-clip-1" \
  -d '{
    "model": "gemini-omni-flash-1.1",
    "prompt": "Slow push-in on the carved turkey, steam rising, warm kitchen light",
    "duration": 6,
    "resolution": "720p",
    "aspect_ratio": "9:16",
    "generate_audio": false,
    "frame_images": [{
      "type": "image_url",
      "image_url": { "url": "https://media.sume.com/artifacts/artf_demo/turkey.jpg" },
      "frame_type": "first_frame"
    }]
  }'

Voice, bed and the facts on screen

The script has three parts: what the meal is, how many it feeds, and how to order. About 380 characters is $0.02 at $0.0475 per 1,000 characters. Make it a wav, set it as the render's audio.url and put the two clips in video[] at 0 and 6 seconds. Add a warm, slow bed as soundtrack with duck_db of about 10 so the voice stays on top. The bed is $0.125.

Then run one caption job with cues: the price per meal and the feeding size from 0 to 6 seconds, and 'Order by Sunday, November 22 / Pickup Wednesday, November 25' from 6 to 12. Replace those dates with yours. The caption job is $0.20 for a video up to 60 seconds, and the cue text is a plain string, so the date cannot be mangled by a model.

Checks before you post

Read the captions against your order sheet. Check that the food on screen matches the menu: if a clip shows a side you do not sell, change the photo or the prompt. If you list allergens or dietary notes, put them in the caption text and in your order page, not only in the voice. Keep the final file under any size limit of the platform where you post it, and check the vertical crop on a real phone.

Retakes are cheap and separate. A new clip is $0.75, a new voice is $0.02, and any change means a re-render at $0.10 and re-captioning at $0.20.

Make the next one from the same files

Keep the two clips, the voice and the bed. For the Christmas takeout menu you need new photos and new cues, and everything else is the same. A new pair of clips is $1.50, a new voice is a few cents, and the render and captions are $0.10 and $0.20. Store each finished render's URL with the date, the menu and the cue text so you can see what you ran last year and whether the order-by date moved.

If you serve more than one site, keep one master and change only the caption cues per site: the address, the pickup hours and the phone number. That is one caption job at $0.20 per location, on the same rendered file.

Cost

• The same ad with 1080p clips: $1.13 each, so $0.76 more in total.
• Draft at 360p first: $0.23 per clip.