Photographer holiday mini-session reel: 4 clips, music, $1.43

To promote holiday mini-sessions, turn four of your own sample portraits into 4-second clips, put a music bed under them, and burn the dates and slot length on as captions. On Sume that is four Wan 3.0 image-to-video jobs at 480p, one Music 1.0 track, one Timeline render and one caption job: $1.425 (about $1.43) for a 16-second 9:16 reel, before you re-run anything.

Mini-session ads have one job: make a viewer picture themselves in the frame and see a date. The photos do the first, the captions the second. Everything below uses documented Sume routes, and the only inputs you supply are four public HTTPS photo URLs (use portraits whose subjects agreed to appear in your promotion).

What does the reel cost, step by step?

Wan 3.0 bills 2 to 30 seconds per job, and the Video Router page says Sume bills list x 1.25. The per-second figures below are the 480p rate from the price table in the Sume repository, so confirm them against GET /v1/videos/models before you budget. Send resolution explicitly: when a Wan 3.0 request omits it, the default is 1080p, which is four times the 480p rate.

Step 1: one clip per photo

Send each photo as a first_frame in frame_images, with resolution set to 480p, duration 4 and aspect_ratio 9:16 (Wan 3.0 accepts 9:16 on image-to-video). Ask for small, believable motion: a scarf lifting, a child turning toward the camera. Keep each prompt about the photo you actually sent. When a job completes, read the video artifact URL (https://media.sume.com/artifacts/...) from GET /v1/jobs/:id/result, where the Jobs and results page shows result.artifacts[], and keep that Sume URL, not any provider URL. Timeline needs these artifact URLs in step 3.

Step 2 and 3: a bed, then one render

Music 1.0 has no duration field and rejects duration or duration_seconds, so put the length in the prompt (for example "a 30-second track") and ask for "Instrumental, no vocals." Timeline then uses the track as the audio spine and cuts it at audio.duration_seconds, so ask for a track longer than the reel; I found no documented behavior for a track shorter than duration_seconds. The Timeline plan route is unbilled, so run POST /v1/timeline-1.0/plan first to see billable_minutes and the estimate. Timeline needs every URL to be a media.sume.com artifact of your workspace, which is what the generation jobs return.

curl -X POST https://api.sume.com/v1/timeline-1.0/render \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: minisession-reel-001" \
  -d '{
    "audio": {
      "url": "https://media.sume.com/artifacts/artf_music/track.mp3",
      "duration_seconds": 16
    },
    "video": [
      { "source_url": "https://media.sume.com/artifacts/artf_a/clip.mp4", "start": 0, "duration": 4 },
      { "source_url": "https://media.sume.com/artifacts/artf_b/clip.mp4", "start": 4, "duration": 4,
        "transition": { "type": "fade", "duration": 0.25 } },
      { "source_url": "https://media.sume.com/artifacts/artf_c/clip.mp4", "start": 8, "duration": 4,
        "transition": { "type": "fade", "duration": 0.25 } },
      { "source_url": "https://media.sume.com/artifacts/artf_d/clip.mp4", "start": 12, "duration": 4,
        "transition": { "type": "fade", "duration": 0.25 } }
    ],
    "output": { "fade_out_seconds": 1 }
  }'

Step 4: the captions that sell the slot

The mixed track has music but no speech, so use authored cues instead of speech-to-text. Each cue is text, start and end in seconds, and a cue-only job never runs transcription. Four cues map to the four clips:

• 0-4 s: "Holiday mini-sessions"
• 4-8 s: "20 minutes, 6 edited photos"
• 8-12 s: "Saturdays and Sundays in December"
• 12-16 s: "Book the link in bio"

What to check before you post

Open the render once with the sound on and read each cue against the final clip. If a date changes, the music and clips stay valid: only the caption job re-runs, for $0.20, on the clean Timeline output (keep that file and burn captions onto it each time). If you want three booking-link variants, that is three caption jobs on one render, not three new renders.