Contents
0%Hey!
Seedance 2.0 is finally available on all the major AI video platforms. If you want to make UGC for your brand, it is the best model for it right now.
In this post I'm walking through 7 UGC use cases I've been running for real ecom brands. Every prompt, every input, every trick I learned. Copy this straight onto your own product.
See It In Action
Before I break down the formats, look at this. One product image, one prompt. Seedance improvised the whole scene, the cuts, the voiceover, and even the dog's behaviour around the product.
Zesty Paws 8-in-1 Multivitamin BitesZesty Paws 8-in-1 Multivitamin BitesThe dog sniff hook is executed perfectly, the scene variety is wild, and the product consistency is spot on. The label, the orange colour, even the shape of the treat. All preserved.
This is what Seedance does when you give it creative freedom. One prompt, full video.
Use Case 1: Full Story UGC
Full Story UGC is where you give Seedance a product image and a scripted voiceover, then let the model improvise the scenes, the cuts, and the B-roll. Best for when you want to move fast and let the model do the creative work.
The Zesty Paws example above is the format. Here is the exact prompt I used.
No captions.
🎥 [Start with upbeat music. Close-up on a dog nose sniffing, then a hand twisting open the orange Zesty Paws jar]
YOU (laughing): "Okay so my dog literally RUNS to the kitchen now every morning… because of these! 🐾✨"
🎥 [Show you holding up the jar, dog sitting & staring at it, tail going crazy]
YOU: "They're Zesty Paws 8-in-1 Multivitamin Bites, joints, coat, digestion, immune support, all in one little chew he thinks is a treat! 🦴🧡"
🎥 [Quick flash of dog running at the park, zooming in the yard, coat shining in the sun]
YOU: "No hiding pills in peanut butter, no powder he won't eat. Just one bite a day and he's thriving! 😍"
🎥 [Show jar on counter + dog licking your face]
YOU (final): "Zesty Paws, the easiest part of being a dog mom! 💛🐶"
The structure is simple. Scene direction in brackets, spoken line right after. You can swap the product and rewrite the lines and it works for pretty much any niche.
The tradeoff: the looser your prompt, the more room Seedance has to go in a direction you don't expect. So this use case is great for speed, not for pixel-perfect control.
For more control, you go talking head.
Use Case 2: Talking Head with Avatar Control
Talking head is the classic UGC format. A real-looking person talking to camera, holding your product. This is what most ecom brands want.
The first problem you hit: Seedance blocks close-up face uploads. So if you want a specific looking actor, you need a workaround.
The Blur Face Trick
The fix is stupidly simple. Blur the face before you upload.
Add blur on top of the image in Canva
Open the actor photo in Canva. Drop a blur effect on the face. Just enough so the features are suggested but not readable. Export and use that as your reference input.
Or just use Starpop's Add Actor feature
Add Actor handles it in the background for you. You upload the face, Starpop blurs it automatically and passes it through to Seedance as a valid reference. Same result, no manual editing.
The Talking Head Generation
Here is the result using CeraVe as the product. One blurred actor image, one product image, and a structured prompt about what she says and when she brings the product into frame.
Blurred actor reference
CeraVe moisturizing lotionBlurred actor referenceCeraVe moisturizing lotionThe prompt:
Use @image[1] as a start frame and make her talk about how her skin was really breaking out a month ago but then she started using the CeraVe skin moisturizer and now her skin is way better. She brings the @image[2] into frame after a few seconds showing it.
The product consistency is great. The bottle shows up correctly, the label is readable, and the tone feels natural. The small text on the bottle distorts slightly if you look closely, which is just a reality of AI video for now.
Use Case 3: Video Extension (Seedance-Only)
This one is the biggest unlock of Seedance 2.0 and no other model does it well.
Once you have a talking head clip you like, you can pass it back to Seedance as a video reference and just ask it to continue. Same actor, same voice, same product, same environment. Carried across cleanly.
That means you can make a full 30-second UGC ad in 2 generations instead of juggling multiple tools.
The prompt for the extension:
Continue @video[1] by making her continue saying "And the crazy thing is it's like $15 at the drugstore. I used to spend so much money on expensive stuff that did nothing. This literally fixed my moisture barrier in like two weeks. If your skin is freaking out, just try it."
You generate 15 seconds. Extend 15 more. Done.
Use Case 4: Green Screen UGC
Green screen is crushing on TikTok right now. An actor talking to camera with a visual, a screenshot, or a study sitting behind them. Works especially well for products with strong claims or research you want to point at.
Research paper background referenceResearch paper background referenceThe prompt:
Green screen style video with an Asian doctor UGC actor with as a background reference @image[1].
[Hook, actor points to research paper in background] Actor: "Check this out, a huge study found people with obesity are 70% more likely to get seriously sick from infections like the flu. Wild, right?"
[Transition, lean in, conversational] Actor: "It's not just about weight, it's about energy, focus, and how your body handles stress every day."
[Solution tease, friendly] Actor: "Simple stuff like supporting your body with adaptogens like ashwagandha can be a huge help in keeping your energy and immune system on track."
One thing to watch: Seedance sometimes mispronounces complex words. Ashwagandha came out slightly off in my generation. Worth doing a pronunciation pass before locking the voice.
Aside from that, the format is nailed in one shot. The background is clean, the actor looks real, and the format is exactly the TikTok style you see crushing right now.
Use Case 5: Unboxing
Unboxing is where Omni-Reference mode really pays off. The more reference images you give Seedance, the more it actually uses them. For a SKIMS care package unboxing I gave it 6 images: packaging, product, different environments.
SKIMS reference 1
SKIMS reference 2
SKIMS reference 3
SKIMS reference 4
SKIMS reference 5
SKIMS reference 6SKIMS reference 1SKIMS reference 2SKIMS reference 3SKIMS reference 4SKIMS reference 5SKIMS reference 6The prompt:
Please create an unboxing video of @image[1]. Reference images @image[2] @image[3] @image[4] @image[5] @image[6] with no music, voiceover or anything, just unboxing and girl opening packages, we don't see girl. Full video focused on unboxing.
The big lesson here: Seedance improvises well but it improvises even better when you feed it specific inputs. If you want a usable unboxing asset in 15 seconds, give it the actual packaging, the actual product, and the actual environment references.
Use Case 6: Lip Syncing Your Own Voice
This is huge for founder-led marketing. You record your own voice, upload the audio, and Seedance generates an AI influencer that lip syncs to it. Your exact words, your tone, your script.
Nocco canNocco canThe prompt:
Make a muscular man in a busy gym hold the @image[1] product and say @audio[1].
The lip sync accuracy is really good. The timing is tight and it matched every word. One thing I noticed: the audio environment matters. I recorded in a quiet room but the scene is a busy gym, so the voice feels a little disconnected from the setting.
If you're recording audio for a gym scene, try to record in a space that matches. Or do some post-processing on the audio to add ambience. Small thing, but it makes the result feel a lot more real.
Use Case 7: Cinematic Ad (Bonus)
This is the format I didn't expect to work but completely did. You can put yourself in any scene you can describe, with full face consistency across a 15 second clip.
I wrote a detailed second-by-second scene breakdown, gave Seedance a photo of my face and a watch product shot, and asked for an Alice in Wonderland wedding sequence.
David face reference
Watch productDavid face referenceWatch productFace consistency is perfect throughout the whole clip. The watch closeup at the end is clean.
This format probably isn't your first move as an ecom brand, but it is a great top-of-funnel play for something that feels like a film rather than an ad. The full prompt is in the free Notion doc (link below) because it's about 400 words long.
Tips & Tricks
After running all 7 of these formats, here are the three things that will save you the most time.
Give Omni-Reference as much context as you can
Seedance improvises well. But the more reference images you give it, the less you leave up to chance. The SKIMS unboxing used 6 reference images and used every single one. Don't skip inputs.
Use the blur trick when you need a specific actor
Close-up face uploads get blocked. A blurred version of the same image works. In Starpop this is handled for you via Add Actor, but you can do it yourself in Canva in 10 seconds.
Extend instead of regenerating
If you need 30 seconds of a talking head, don't regenerate twice and cut. Generate once, then feed the clip back in as a video reference and ask Seedance to continue. This is the biggest advantage of Seedance 2.0 over every other model right now.
Next Steps
That's 7 UGC formats you can run on a real product today. Pick the one closest to what your brand needs and copy the prompt.
- Free Notion doc with every prompt and input is pinned in the Starpop Discord. Including the full cinematic ad prompt that didn't fit in this post.
- Seedance 2.0 templates for each of these formats are rolling into Starpop this week.
Drop your product image in the Starpop Discord and I'll run it through Seedance 2.0 for you, completely free. A lot of the examples above came from people in the community who did exactly that.
