seedance-video-ad-generator: video ad prompts
What it does
Section titled “What it does”seedance-video-ad-generator writes production-ready AI video ad prompts for direct response. Give it a brand and an ad idea and it returns shot-by-shot prompts optimized for Seedance 2.0, ready to paste into Dreamina.
It handles three constructs: a single shot (5-8s, one beat, for hooks), a multi-shot scene (2-3 beats in one generation, for demos and transformations), and a multi-clip chain (three 8-second generations stitched together in post for a longer arc).
Every prompt follows a locked formula: camera, subject, action, environment, lighting, style, constraints. Seedance weights what it reads first, so camera always goes first.
When to reach for it
Section titled “When to reach for it”- You want AI UGC-style video ads for Meta Reels, TikTok, or YouTube Shorts.
- You have a demo, transformation, or hook concept to turn into a shootable prompt.
- You’re building a multi-clip narrative: problem, solution, CTA across three clips.
- You want to test 3-5 hook variations fast, or build an ad around a product photo.
Where it runs
Section titled “Where it runs”Prompt-writing and shot planning work in both Cowork and the desktop app: you get paste-ready prompts either way. Generating the actual video is a separate step. Seedance is an outside service (Dreamina / ByteDance) that Cowork’s sandbox blocks today, so finished files need the desktop app with API keys configured (see API keys setup).
How to run it
Section titled “How to run it”Describe the concept and name the brand, with whatever detail you have:
“seedance-video-ad-generator: product demo for examplebrand.com, woman applies the serum, before-and-after reveal, 15s, Reels”
Or have it run the whole pipeline from scratch:
“Seedance ads for examplebrand.com, full guided build: research, personas, then 5 hook prompts”
What it needs from you
Section titled “What it needs from you”- A brand URL or a clear product description. A URL is better: it pulls real colors, tone, and product details.
- The ad construct: single-shot, multi-shot, or multi-clip chain. (It asks if you don’t say.)
- A product photo if the ad shows your branded product. Text alone produces a generic lookalike. You attach it at generation time in Dreamina, and the skill flags exactly when.
- (Optional) Which placement: Reels, TikTok, or Shorts. Default is 9:16.
It works best after you’ve onboarded your brand, but it can run cold on any URL.
What you get back
Section titled “What you get back”For each prompt, a formatted block with three parts:
- Model spec: duration, aspect ratio (9:16 default), resolution, audio on/off, and any reference image to attach.
- Prompt body: the full camera-to-constraints formula, ready to paste. Single-shot prompts stay under 70 words. Multi-shot prompts use
Shot 1 / Shot 2 / Shot 3or a[0s] / [3s] / [6s]timeline. - Post-production notes: what overlays or voiceover to add after you generate, plus how the clip chains in a multi-clip arc.
For variations, each differs on one variable (camera angle, lighting, action beat) so you can isolate what drives performance.
A worked example
Section titled “A worked example”You: “Seedance prompts for examplebrand.com, a magnesium sleep supplement. Single-shot UGC hook, woman waking up refreshed.”
seedance-video-ad-generator: (after loading brand and persona context)
PROMPT 1: morning refresh hook
Model spec: Seedance 2.0 Standard, 8s, 9:16, 1080p, audio on. Attach the product bottle photo as a reference at generation time.
Prompt body: Handheld close-up. Woman in her late 30s, natural makeup, linen bedding, stretches overhead and exhales slowly. Bedroom morning, soft window light from camera left. Authentic UGC phone-shot aesthetic, photorealistic.
Post-production notes: Caption in safe zone, “Woke up before my alarm for the first time in years.” Chain with a product-detail clip for a 3-clip arc.
It opens with camera and pushes captions to post, so the model never renders text.
Tips & gotchas
Section titled “Tips & gotchas”- Captions are mandatory, but they go in post. 85%+ of Reels and TikTok plays are muted, so every ad needs burned-in captions after you generate. The skill tells you what to put and where.
- The hook must land in the first 1-2 seconds. If those two seconds could belong to any video in the feed, the ad won’t work, so the skill builds the pattern interrupt into the shot structure.
- One camera move per shot, and specific lighting. Two moves (“push in while panning right”) cause jitter, so the formula enforces one. “Warm light” is not an instruction; “practical window light, overcast diffuse” is.
- Branded products need a reference image. Attach a real photo at generation time in Dreamina (the skill flags this in the model spec). Without it, you get a lookalike with wrong colors and no label.
- UGC beats polished by default on paid social: handheld, natural light, authentic setting. Say so if your brand is luxury. In multi-clip arcs, the character anchor phrase repeats word-for-word across all three prompts, since Seedance relies on it for consistency.
Related skills
Section titled “Related skills”These prompts feed directly into the video generation step. Pair it with apollo (to ground the creative in real buyer language) and hermes (to write the on-screen copy and voiceover for post). For motion-graphic or text-driven ads where you don’t want AI footage, reach for motion-video instead.