Hand-drawing storyboards is the old way. Today you generate them with image models like GPT-Image-2, drop the result into Seedance 2.0 as a reference, and the storyboard becomes video. You don't need to draw — and Seedance handles camera moves and editing on top.
scenic.sh shows the whole chain. The result video, the Seedance prompt that produced it, the storyboard image used as reference, and the image prompt that generated that storyboard. Copy any layer and use it.
The 5 prompts below each use storyboards differently — 6-shot shorts, 4x4 grids, one-line commercial boards. Compare the structures.
1. Boy and the Kite — 6-Shot Emotional Short
See full prompt & storyboard →
"Shot 1: A young boy stands in a grassy field, looking up at a red kite tangled in a tree. Shot 2: Close-up of the boy holding the kite, smiling as he comes up with a plan. Shot 3: Indoors, he carefully repairs the torn kite with tape at a wooden table. Shot 4: Back outside on a hill, he runs and launches the kite into the sky..."
Why this works: The most basic and the most powerful storyboard format. Number the shots 1 through 6, and put location, action, emotion in one sentence per shot. Seedance reads this structure and decides scene transitions and camera moves automatically. Proof that with a clear narrative — no fancy technical jargon required — you get a cinematic result. The reference storyboard here was generated as a 6-frame sequence with GPT-Image-2 — and on the scenic page you can see the image prompt that produced it.
2. Retro Sci-Fi — 4-Scene Action Sequence
See full prompt & storyboard →
"Scene 1 — Action: Vera leaps onto the hood of her patrol car, racking the slide of her raygun as a saucer descends rapidly. Camera: Low angle, tracking shot following her movement, quick zoom to close-up on her determined expression. Lighting: High contrast, dramatic rim lighting with harsh amber highlights..."
Why this works: Each scene is broken into Action / Camera / Lighting / Sound / Dialogue. This is the same structure as a real film continuity sheet. Seedance recognizes these categories and processes them independently. Sound directives in particular pull through into effects and ambient audio. Specify camera technique — "low angle, tracking shot" — and Seedance hits the exact angle. The storyboard itself was generated as 4 separate scenes with an image model, then fed to Seedance.
3. Commercial Storyboard — One Line Is Enough
See full prompt & storyboard →
"Create a commercial ad from the storyboard @[image1]"
Why this works: That's the entire prompt. One line. The real work is in the image (@image1). Generate a clean commercial board with GPT-Image-2 and Seedance reads layout, sequence, and scene composition straight from the image. The single genre keyword "commercial ad" determines editing rhythm, color, and pacing. The clearest proof that when the storyboard is good, the prompt should be short.
4. Bird and Boy — Emotion-Driven Animation
See full prompt & storyboard →
"Use image as storyboard, created dynamic animated sequence, with emotions, enlightening music score. Don't show the text, it's just indication for the story. Create dynamic camera angle, 3D feel. Use sound of the bird and voice for the boy."
Why this works: "Don't show the text, it's just indication for the story" — that line is the key. When GPT-Image-2 generates a storyboard, it often includes dialogue and direction text right inside the panels — this instruction tells Seedance not to render that text on screen, but to use it as direction reference only. "3D feel" is a short but effective keyword for camera moves with depth.
5. GRWM Fashion — 4x4 Grid Storyboard
See full prompt & storyboard →
"REFERENCE IMAGE: Use the provided 4×4 male storyboard collage as visual guide for framing, pacing, outfit flow, and transitions. CONCEPT: Get Ready With Me — Modern Minimal Menswear TIMELINE: 0:00–0:04 Dark fitted tee, ivory tailored trousers, clean tuck-in adjustment, mirror check..."
Why this works: Uses a 4x4 collage as the storyboard while pinning timing down to the second with a TIMELINE. An image model generates 16 panels in one clean layout; the text declares "when each appears." Calling out edit techniques in the TRANSITIONS section — "match cuts (tuck → belt → blazer)" — pushes the polish further. Especially well-suited for short-form content like GRWM.
Summary
| Method | Prompt Length | Best For |
|---|---|---|
| Numbered shots | Medium | Emotional shorts, narrative-led |
| Scene + Camera/Lighting/Sound | Long | Action, cinematic |
| One line + image | Short | When the storyboard is highly polished |
| Image + emotion/sound keywords | Medium | Animation, emotion-driven |
| Grid + timeline | Long | Short-form, fashion, GRWM |
The higher the storyboard's quality, the shorter the Seedance prompt can be. Image models like GPT-Image-2 produce great boards now, so the highest-leverage move is making a strong storyboard up front.
Each prompt page on scenic.sh gives you the result video + Seedance prompt + reference storyboard + the image prompt that made the storyboard — all copyable.
