I’ll break down the reference ad’s hook, visuals, and pacing, then adapt it for StarDust using your attached packaging as the visual source of truth.
The reference is a 101-second elevator comedy with 28 shots: two women gossip about a gym-goer in Korean, discover she understands them, then ask about her routine. The product reveal arrives late, followed by an embarrassed buying moment and an English closing voiceover.
I’ll keep the elevator setting, awkward-comedy hook, reaction shots, and pacing, replacing Primal Queen with StarDust.
For the product section:
- Reveal your cream-and-coral StarDust sachets, matching the attached image.
- Replace the beef-organ exchange with a citrus-electrolyte exchange.
- Adapt the closing product line to StarDust while keeping the training-and-discipline setup.
One choice before I adapt the full script: keep the original Korean dialogue with English subtitles and an English closing voiceover, or make it all English?
I recommend keeping the bilingual version—the “you understood us?” reveal is the joke that makes this ad work.
I’ll keep the Korean dialogue, English subtitles, and English closing voiceover. I’m checking the exact lines around the product reveal so only the StarDust-specific exchange changes.
Locked: Korean dialogue, English subtitles, and an English closing voiceover, keeping the reference’s elevator comedy and roughly 101-second pacing.
I saved the complete shot-by-shot adaptation here:
StarDust — Elevator Ad | Korean Dialogue & English Subtitles
The main change is the product-reveal exchange:
- Friend: “전해질이요?”
“Electrolytes?” - Gym-goer: “네, 스타더스트. 물에 타 마시는 시트러스 전해질이에요.”
“Yes, StarDust. Citrus electrolytes you mix into water.”
The buying line becomes “I'm buying StarDust.” The English closing ends:
“Early mornings, real discipline, and one electrolyte blend I swear by. StarDust.”
This replaces the original beef-organ joke with a straightforward electrolyte reveal. I recommend this version for a close adaptation. Approve the script, and I’ll move into the storyboard and shot recreation.
Agreed. I’ll use Seedance 2.5 at 480p and plan the ad in multi-shot scenes, with cuts and reaction beats specified inside each generation.
I’ll first lock references for the three women, the elevator, and StarDust. Then I’ll group the 28 source shots around complete dialogue exchanges so each batch preserves the scene’s geography and performances.
Seedance 2.5 supports 480p, up to 30 seconds per generation, and multiple image references. I’ll use five batches, keeping the “Obviously surgery / You can’t get that from working out” exchange together rather than splitting it between generations.
| Batch | Source shots | Planned length | Dramatic beat |
|---|---|---|---|
| A | 1–6 | 30s | Entrance, gossip, and the unexpected “Thank you” |
| B | 7–13 | 20s | Korean reveal, shocked reactions, apology |
| C | 14–20 | 25s | Training routine, StarDust reveal, exit |
| D | 21–25 | 16s | Embarrassment and the buying punchline |
| E | 26–28 | Measured English VO | Knowing smile, then StarDust packshot |
Continuity rules: the friends keep their left/right positions; the gym-goer stays nearer the doors; wardrobe, bags, drinks, lighting, and screen direction stay locked. Each batch will use the same canonical references—not the previous batch’s output.
Shot plan
Timing below follows the reference; native-dialogue timing may shift slightly. G = gym-goer, A/B = friends, E = elevator, P = your StarDust product image. Character and environment sheets are next to be made.
| Shot | Visual | Timing / duration | Script / Audio | References | Transition-in |
|---|---|---|---|---|---|
| 1 | Doors open; G enters with her duffel | 0–2.97 / 2.97s | LIPSYNC: both “Hi” greetings | G, A, B, E | Opening |
| 2 | Interior wide; G presses a button and faces forward | 2.97–6.43 / 3.46s | Silent picture | G, A, B, E | Cut |
| 3 | Rear full-body view of G facing the doors | 6.43–11.83 / 5.40s | A speaks off camera: admiration | G, E | Cut |
| 4 | Over G’s shoulder; A leans toward B | 11.83–22.73 / 10.90s | LIPSYNC A: jealousy and surgery question | G, A, B, E | Cut |
| 5 | Tight shot of B answering with certainty | 22.73–26 / 3.27s | LIPSYNC B: “Obviously surgery” | B, E | Cut |
| 6 | G registers the remark and begins turning | 26–29.57 / 3.57s | B finishes off camera; LIPSYNC G: thanks | G, E | Cut |
| 7 | G faces the friends with a small smile | 29.57–33.30 / 3.73s | LIPSYNC G: all natural | G, E | Cut |
| 8 | Friends freeze, eyes widening | 33.30–34.17 / 0.87s | Silent reaction | A, B, E | Cut |
| 9 | Close-up of A looking toward G | 34.17–37.33 / 3.16s | LIPSYNC A: Korean question | A, E | Cut |
| 10 | Close-up of G answering calmly | 37.33–42.17 / 4.84s | LIPSYNC G: Seoul explanation | G, E | Cut |
| 11 | Friends bow apologetically, drinks held close | 42.17–44.17 / 2s | LIPSYNC A: apology | A, B, E | Cut |
| 12 | G smiles and waves off the apology | 44.17–46.30 / 2.13s | LIPSYNC G: compliment | G, E | Cut |
| 13 | A straightens and asks earnestly | 46.30–49.47 / 3.17s | LIPSYNC A: routine question | A, B, E | Cut |
| 14 | Medium close-up of G describing her routine | 49.47–53.70 / 4.23s | LIPSYNC G: weights, protein, sleep | G, E | Cut |
| 15 | A asks a follow-up over G’s shoulder | 53.70–56.90 / 3.20s | LIPSYNC A: supplements | G, A, E | Cut |
| 16 | G nods and lowers her gaze to her bag | 56.90–59.10 / 2.20s | LIPSYNC G: “Of course” | G, E | Cut |
| 17 | Insert: G retrieves a StarDust sachet | 59.10–61.37 / 2.27s | Silent product beat | G, P | Cut |
| 18 | Friends look down at the sachet | 61.37–65.20 / 3.83s | LIPSYNC A: “Electrolytes?” | A, B, E, P | Cut |
| 19 | Side three-shot; G presents and passes the sachet | 65.20–69.73 / 4.53s | LIPSYNC G: StarDust explanation | G, A, B, E, P | Cut |
| 20 | G exits, turns back, and waves | 69.73–73.83 / 4.10s | LIPSYNC G: goodbye | G, E | Cut |
| 21 | Two-shot; A holds StarDust and winces | 73.83–77 / 3.17s | LIPSYNC A: embarrassment | A, B, E, P | Cut |
| 22 | Tight reaction of A avoiding eye contact | 77–77.97 / 0.97s | Silent reaction | A, E, P | Cut |
| 23 | B gives A an awkward sideways glance | 77.97–80.43 / 2.46s | A speaks off camera: elevator line | A, B, E | Cut |
| 24 | Wider two-shot; B reaches for her phone | 80.43–85.40 / 4.97s | Silent pause | A, B, E, P | Cut |
| 25 | Close-up of B looking at her phone | 85.40–88.93 / 3.53s | LIPSYNC B: buying StarDust | B, E | Cut |
| 26 | G stands outside the elevator, knowingly smiling | 88.93–91.43 / 2.50s* | VOICEOVER: “They thought…” | G, E | Cut |
| 27 | Closer on G’s smirk | 91.43–93.97 / 2.54s* | VOICEOVER: “And I did…” | G, E | Cut |
| 28 | StarDust box and sachet beside a bottle and towels | 93.97–101.13 / 7.16s* | VOICEOVER: discipline and StarDust close | P, E | Cut |
*The closing shots will be fitted to the measured English voiceover. English subtitles will be added in the edit, not generated into the footage.
I’ll now build the reusable character and elevator references and record the closing voiceover before creating the shot guides.*
The character sheets keep the three women visually distinct, and the elevator sheet captures the lighting, metal walls, and doorway layout. I spotted a prop error in Friend A’s sheet—her coffee cup is missing from the front view—so I’ll correct that before using it.
Your supplied StarDust image is confirmed as the product reference; I won’t regenerate its packaging design.
The corrected Friend A sheet now includes her coffee and takeout bag consistently. The English closing voiceover is clean and measures 9.53 seconds, so its three shots will use that timing rather than stretching the recording.
Next I’ll make the shot images that guide each batch. The videos will still be generated as five multi-shot scenes; the individual images give Seedance a concrete composition for every cut.
The first 30-second test produced all six planned shots, and the visible dialogue is lip-synced. I also checked two words the automatic transcript misread; the audio review found the intended Korean lines clearly spoken.
There is one continuity issue: Friend A briefly acquires a phone in the entrance shot, then holds coffee inside. I’ll make her coffee explicit in that opening reference and rerun the scene before presenting it as the multi-shot test.
The rerun fixed the phone, but its dialogue timing was worse. I’ll keep the stronger original performance and replace only its opening shot with the corrected take. That fixes the prop continuity without sacrificing the better lip-sync.
I’ll add the approved English subtitles to this opening-scene preview so you can judge the multi-shot approach before we spend on the remaining four scenes.
The 30-second opening-scene preview is ready, generated with Seedance 2.5 at 480p.
- Six planned shots within one multi-shot generation.
- Locked character and elevator references.
- Korean dialogue with English subtitles.
- Corrected the opening’s coffee/phone continuity issue using the stronger sections of the two takes.
The rendered preview passed technical checks: no black or frozen frames and no subtitle-speed warnings.
This is the opening scene, not the full ad. The remaining four scenes are planned; the English closing voiceover is also ready.
I recommend reviewing this preview’s performances and cut rhythm before generating the rest. Does this hit the cinematic drama-ad feel you want?
You’re right—she should still be facing the elevator doors, unaware of the gossip. I’ll correct this frame so only the back of her head is visible, keeping everything else unchanged. Her turn toward the girls belongs at the later “Thank you” reveal.
The first corrections still leave her profile visible. I’ll use a different image model to make the head turn unambiguous, while preserving the girls, props, and framing.
Corrected—only the back of her head is visible now, with no glance toward the girls. She’ll stay facing the doors throughout the gossip and turn only at the “Thank you” reveal.
This fixes the shot reference; the video preview hasn’t been updated yet.
I’ll finish the full ad now: update the opening with her head facing the doors, generate the remaining multi-shot scenes in Seedance 2.5 at 480p, then assemble them with English subtitles and the approved closing voiceover.
I’ll keep the dialogue unchanged and check the StarDust packaging, speaker assignments, and continuity before rendering.
The updated opening now keeps her facing the doors throughout the gossip; she turns only for the “Thank you” reveal. The visible speakers’ lip-sync also checked out.
The StarDust shot guides are ready. I’m generating the product reveal, the friends’ embarrassed exchange, and the closing sequence next, then timing the subtitles to the finished dialogue.
All five scenes are generated. The audio check confirmed that the buying punchline is present—the automatic transcript had missed it—and that the StarDust explanation is clearly pronounced.
I’m assembling the final cut now. I’ll remove an added “um” before “Of course,” fit the closing shots to the approved voiceover, and keep the product label sharp during the final push-in.
The full timeline is assembled at about 100 seconds:
| Shots | Sequence | Final timing |
|---|---|---|
| 1–6 | Corrected entrance and gossip | 0:00–0:30 |
| 7–13 | Korean reveal and apology | 0:30–0:50 |
| 14–20 | Routine, StarDust reveal, exit | 0:50–1:14 |
| 21–25 | Embarrassment and buying punchline | 1:14–1:30 |
| 26–28 | English voiceover and product close | 1:30–1:40 |
I’m rendering the complete ad and checking subtitle readability, audio levels, and the final joins.
Finished—the full 1:40 StarDust ad is rendered.
- Korean dialogue with English subtitles
- Corrected head direction during the gossip
- English closing voiceover, subtle music, and StarDust product close
- Generated with Seedance 2.5 at 480p, exported vertically at 1080 × 1920
Technical checks passed: no black or frozen frames, readable subtitles, and audio within target levels.
Yes—I’ll crop that gossip shot tighter around the two friends so her face is out of frame, with just a sliver of shoulder at the edge. No regeneration; the dialogue and timing will stay unchanged.
Done—cropped the entire gossip two-shot (0:11.7–0:22.7) around the two friends. Her head is out of frame, with only a little shoulder visible.
No regeneration; dialogue and overall timing are unchanged. The tighter crop is slightly softer because the source footage is 480p.
