Cohort course waitlist teaser: two Wan clips and dates for $1.05

Build a 10-second 9:16 waitlist teaser for a cohort course on Sume: two Wan 3.0 B-roll clips, Music 1.0, Timeline and date captions for $1.05.

4 min readSume
All posts

A waitlist teaser for a cohort course needs three facts on screen: what it teaches, when the cohort starts, and when enrollment closes. Two 5-second Wan 3.0 B-roll clips at 480p, a Music 1.0 bed, a Timeline render and one caption job make a 10-second 9:16 teaser for $1.05 on Sume, with no avatar and no filming.

A cohort differs from an evergreen course in one way that matters for video: the dates expire. The teaser you post for the waitlist is wrong on the day the cart opens, and wrong again when the cart closes. So build the footage once and treat the captions as the part you will re-run.

Cost of the 10-second teaser

Prices are Sume's per-job and per-second rates; the only variable is how many clips you generate.

Sume docs read 2026-10-10: Video Generation, Music 1.0, Timeline 1.0, Video captions. Wan 3.0 480p rate (list x 1.25) from the repository price table; confirm in GET /v1/videos/models.
StepBasisCost
Two B-roll clips2 x 5 s x $0.0625/s, 480p$0.625
Music bedFixed per generation$0.125
Timeline render10 s rounds up to 1 minute$0.10
Date captions1 job$0.20
Teaser total$1.05
Each later date changeRe-caption the clean render only$0.20

Step by step

Step one is text-to-video for B-roll that suggests the subject without showing a person: a desk, a notebook, a window. Send resolution every time, because a Wan 3.0 request that omits it defaults to 1080p, and send aspect_ratio 9:16. Run two prompts for two different desk scenes so the cut has something to change.

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: cohort-broll-001" \
  -d '{
    "model": "wan-3.0",
    "prompt": "Slow push-in on a tidy desk: open notebook, laptop showing a blank lesson outline, steam from a mug, morning window light",
    "resolution": "480p",
    "duration": 5,
    "aspect_ratio": "9:16"
  }'

Steps two to four

The remaining three steps are one call each, and none of them needs a camera or a person.

  • Music 1.0: a calm 72 BPM piano and soft pad, "a 20-second track", ending with "Instrumental, no vocals." There is no duration parameter, so inspect the result before you rely on its length.
  • Timeline: the music is the audio spine with duration_seconds 10, and two video[] slots at start 0 and start 5 with a short fade into the second. Run /v1/timeline-1.0/plan first; it is unbilled.
  • Captions: send cues on the render. Suggested set: "Cohort 4 starts Jan 12" (0-3 s), "Eight weeks, live on Tuesdays" (3-6 s), "Waitlist gets first access" (6-8 s), "Enrollment closes Jan 5" (8-10 s).

Re-using it when the cart opens

The render has no text on it, so save its URL. When the waitlist closes and the cart opens, run one new caption job with the first cue changed to "Enrollment is open" and the closing date unchanged. When the cart is in its last 48 hours, run another. Each one is $0.20, and none of them touches the footage.

If you already have a talking-head intro, the course pages in this blog cover those (see the related reading below): this post is the no-camera route for the weeks before there is any course content to show.

Sources

Related posts

More in Use cases

All Use cases posts

Written by Sume