Lantern Festival 2027 riddle video: five riddles, one 35-second cut

The Lantern Festival in 2027 falls on 20 February, two weeks after Lunar New Year, and its best-known custom is solving riddles written on lanterns. A riddle video is a natural fit for a short vertical cut: five riddles held for six seconds each, one motion clip of lanterns, one music bed, about $1.22 of Sume usage in total.

Facts about the festival come from Wikipedia's Lantern Festival article, read 2026-10-03: the fifteenth day of the first lunisolar month, 20 February in 2027, with lanterns, riddles and tangyuan (glutinous rice balls) as the main customs, and a link to the end of the New Year celebrations. Sume facts are from the docs linked in Sources.

Why a riddle format works for a business

A riddle has a built-in two-beat structure: ask, then answer. That maps cleanly onto a caption job with two cues per riddle and gives viewers a reason to stay to the end of a six-second slot. A tea shop can ask about leaves, a bakery about the dumpling, a bookshop about a lantern with a poem. The festival supplies the frame; your product is the answer to the last riddle.

The build and the bill

Make the five riddle stills with the Image API (bytedance-seed/seedream-4.5, $0.033 per image in the docs' pricing example). Generate one five-second 9:16 lantern clip with wan-3.0 through POST /v1/videos (accepts 2 to 30 seconds). Put the music bed in as soundtrack and render everything in one Timeline job.

Timing the cues

With five 6-second stills followed by the 5-second clip, the video is 35 seconds long. Put each question at second 0 to 3 of its slot and the answer at second 3 to 6, so cue times are 0 to 3, 3 to 6, 6 to 9, 9 to 12 and so on. A caption job with cues skips speech-to-text, so a silent cut is fine; the docs say a silent clip without cues fails as caption_no_speech.

import os, sys, uuid, requests

def main():
    key = os.environ.get("SUME_API_KEY", "")
    if not key:
        sys.exit("Set SUME_API_KEY first")
    body = {
        "video_url": "https://media.sume.com/artifacts/YOUR_ARTIFACT/clean.mp4",
        "style": "slam",
        "cues": [
            {"text": "What has a face but cannot smile?", "start": 0.0, "end": 3.0},
            {"text": "A clock", "start": 3.0, "end": 6.0},
        ],
    }
    resp = requests.post(
        "https://api.sume.com/v1/video-captions",
        json=body,
        timeout=60,
        headers={"Authorization": "Bearer " + key,
                 "Idempotency-Key": str(uuid.uuid4())},
    )
    print(resp.status_code, resp.json())

if __name__ == "__main__":
    main()

What Sume does and does not do

• Does: make the stills, the clip and the music, assemble them in one MP4 and burn the cues.
• Does not: write riddles that are correct or culturally apt. Write them yourself or have a speaker check them.
• Caution: the caption docs list Latin and Hangul faces only. Keep Chinese characters inside the still, review the image, and use the cues for the English line.

Schedule

Work back from 20 February: riddle copy by mid-January, stills and the clip by the end of January, render and review the first week of February, post on 18 or 19 February so it is live before the day. If you also send a greeting for 6 February, reuse the music bed and the lantern still style so the two feel like one campaign. See the Lunar New Year plan for that first send.