Launch kit: 4 clips, 12 images, a voiceover and a track for $9.98

Four 8-second 1080p clips, twelve 2K images, a 1,200-character voiceover and a track cost $9.982 with Wan 3.0, or $20.4716 with Seedance 2.5 clips.

4 min readSume
All posts

A launch kit of four vertical 8-second Wan 3.0 clips at 1080p, twelve Nano Banana 2.1 images at 2K, a 1,200-character voiceover and one music track costs $9.982 on Sume. Swapping the clips to Seedance 2.5 at 720p makes the same kit $20.4716.

Video is 80 percent of the Wan 3.0 kit, so that is the line to plan around. Images, voiceover and music together are $1.982.

Itemized

Each line comes from the price tool or arithmetic on it. The voiceover is 1,200 characters at $0.0475 per 1,000.

Launch kit, first-take prices (read 2026-10-09)
ItemModel and settingUnit priceCountSubtotal
4 vertical clips, 8 s eachWan 3.0, 1080p$2.004$8.00
12 imagesNano Banana 2.1, 2K$0.1512$1.80
Voiceover, 1,200 charactersTTS, $0.0475 per 1,000$0.0571$0.057
Music bedMusic Router, flat$0.1251$0.125
Kit total$9.982

With a 25 percent retry allowance

Add one retake for every four video clips and one extra image for every four images. That is $2.00 for one more clip and $0.45 for three more images, for a planned spend of $12.432. Keep the voiceover and music at one take each unless the script changes: together they cost $0.182.

Make it a script

List the kit in a manifest and price it before submitting. This keeps the budget in one place and gives each asset a stable idempotency key.

kit = [
    ("video", "wan-3.0", 4, 2.00),
    ("image", "google/nano-banana-2.1", 12, 0.15),
    ("tts", "script-1200", 1, 0.057),
    ("music", "sume/music-auto", 1, 0.125),
]
total = sum(n * unit for _, _, n, unit in kit)
print(f"kit estimate: ${total:.3f}")

Gotchas

  • The unit prices in the manifest are rounded; use the model catalog or the estimate for the exact amount before a large run.
  • Vertical 9:16 on Wan 3.0 does not change the per-second price; on Seedance it can.
  • Voiceover is separate from the clip: check generate_audio in the model's catalog entry before you assume the clips carry sound.

Sources

Related posts

More in Use cases

All Use cases posts

Written by Sume