Coffee ad in one take, pour to sip: Seedance 2.5 prompt and price
A 20-second one-take coffee ad on Seedance 2.5: pour, steam, hand, sip. Prompt wording, why one take suits liquids, and the Sume price at 720p and 1080p.

A coffee ad suits one take because the story is a single motion: pour, steam, hand, sip. On Seedance 2.5 on Sume, a 20-second 16:9 clip costs $11.56 at 720p and $28.44 at 1080p.
ByteDance Seed describes the model as built for continuous one-take shots of up to 30 seconds (read 2026-10-05). A coffee pour needs about a third of that.
Why 20 seconds
Pour 5 seconds, steam and a pause 5, the hand 5, the sip and a closing look 5. At 15 seconds the beats run into one another; at 30 the camera has to invent extra action. 20 is a comfortable middle, and it is a duration inside the 4-30 range on the Sume catalog row.
Prompt wording for liquids
Name the surface the liquid meets and the vessel.
- "A ceramic cup on a wooden counter, side-lit by window light. Espresso pours into steamed milk in a slow thin stream, making a small pattern."
- "Steam rises and moves in the light. The camera holds close and still, a very slow push-in."
- "A hand with no jewelry lifts the cup. Close on the cup lip, a sip, then a small exhale."
- "Ambient cafe sounds, no music, no speech."
Price by length and resolution
Pour-only and full-sequence versions.
| Length | 480p | 720p | 1080p |
|---|---|---|---|
| 10 s | $2.69 | $5.78 | $14.22 |
| 15 s | $4.04 | $8.67 | $21.33 |
| 20 s | $5.38 | $11.56 | $28.44 |
Request
A first-frame photo of the empty cup keeps the cup shape and counter you plan to match.
curl -X POST https://api.sume.com/v1/videos \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: coffee-pour-001" \
-d '{"model":"seedance-2.5","prompt":"Espresso pours into steamed milk in a slow thin stream. Steam rises in window light. Camera holds close and still with a very slow push-in. A hand lifts the cup, a sip, a small exhale. Cafe ambience, no music.","frame_images":[{"type":"image_url","image_url":{"url":"https://example.com/empty-cup.jpg"},"frame_type":"first_frame"}],"duration":20,"resolution":"720p","aspect_ratio":"16:9"}'Variants worth a second render
Once the 20-second pour works, three cheap variants cover most campaigns. A 9:16 version for stories costs the same per clip at a given length, so it is a prompt change plus the aspect ratio, not a new budget line: $11.56 at 720p. A cold-brew version swaps the pour for ice and a slow stir. A morning version changes the light to low and warm and the counter to a kitchen table.
Change only the lighting words or only the vessel between variants. When a coffee prompt changes in five places, the pour can lose the thin steady stream that made the first take work.
Camera and light choices
Side light from a window gives the steam something to catch. Backlight makes the steam glow but can hide the pattern in the milk, so pick one. A very slow push-in suits the pour; a locked camera suits the sip. Avoid orbit moves for a cup, since a round object rotating in the frame invites handle and rim distortion. Keep the counter simple: wood, stone or a plain tray, without props that could pick up text.
What goes wrong and how to check
Liquids are a place where a generated clip can break physics: milk that climbs, a cup that fills while pouring has stopped. Watch the cup level across the pour. If it misbehaves, shorten the pour sentence and say the stream is thin and steady. A 480p retry of this clip is $5.38.
Logos are the other trap. A cup with a brand mark in a reference photo may come out with a distorted mark. If a clean logo matters, leave the cup plain in the clip and add the mark in your editor afterwards. Also decide whether you want a face: the prompt above shows only a hand on purpose, which avoids a person who looks like someone.
Sound is the last choice. Read the seedance-2.5 row in the catalog (GET /v1/videos/models) before you rely on generated audio, and if you want a specific track, add it in the edit.
Sources
Related posts
More in Use cases
- Country and acoustic folk music for a brand ad: a warm brief
Brief an instrumental country-folk bed for a brand ad on Sume: guitars, pace, and a warm ending. A take is $0.125; a 30-second ad renders at $0.10.
- Course syllabus to a 30 s trailer: Wan 3.0 with one beat per week
A six-week syllabus becomes a 30 s Wan 3.0 trailer on Sume: one beat per week, references from course art, and a Timeline join if you want per-week control.
- Course trailer in 4 languages from one video track: $0.57 on Sume
A 45-second course trailer in four languages reuses one video track: each language is about $0.14 of TTS, concat and a one-minute render, $0.57 in all.
- Crowdfunding video music for a 2 to 4 minute pitch without fatigue
A 3-minute pitch video needs music that shifts with the story. Brief three Lyria 3.5 cues at $0.125 each, join with Timeline audio concat at $0.01, then duck.
Written by Sume