Seedance 2.0 Prompts for UGC Ads: 8 Tested Examples

Seedance 2.0 is the right model for UGC ads when motion is what sells the product. If you're shooting a skincare serum, a supplement pouch, or a kitchen gadget and the ad needs to feel like a real person filmed it on an iPhone, Seedance's multi-reference conditioning gets you there faster than any other model right now. The formula: product reference image, environment reference image, an action, an authenticity cue, and a 9:16 aspect ratio. Here are 8 prompts we ran and what each produced.

TL;DR

• Seedance 2.0 wins for UGC when product motion and identity consistency matter more than speed
• Multi-reference (product photo + environment photo) separates a usable clip from a generic one
• Use Kling 3.0 for fast variant volume, Higgsfield for talking heads, Seedance for product-in-use and tactile shots
• Generation time: around 2 min per clip at 9:16, 1080p, 30fps

When to use Seedance 2.0 for UGC

Seedance is not the fastest model and it's not the cheapest. You use it when the product needs to look like the actual product throughout the whole clip. Multi-reference locks the label, shape, and colorway so it stays recognizable from the first frame to the last.

Kling 3.0 is the better call when you need 10 ad variants in an afternoon. It's faster, cheaper, and the variance between runs is useful for A/B hook testing. But on close-up product frames, Kling's label consistency slips. Seedance holds.

Higgsfield Soul 2.0 owns the talking-head format. If your UGC ad is a person-to-camera testimonial, Higgsfield's character conditioning is the right tool. Seedance doesn't have the same face fidelity for human subjects.

Use Seedance for product handling shots, liquid pours, texture reveals, before-after demos, and any format where the product is the subject and needs to stay identifiable frame to frame.

The prompt formula

Every prompt that's worked follows this structure:

Reference image (product) + Reference image (setting) + Action + Authenticity cue + Aspect ratio

The authenticity cue is what keeps the output from looking like a polished commercial. Natural hand tremble, ambient window light, or a slightly casual framing tells the model you want a human-made feel. Skip it and you get something that looks produced. That kills the UGC read.

Keep prompts under 80 words. Seedance reads long prompts literally and outputs get stiff when you over-specify.

8 tested prompts for UGC ad formats

1. Unboxing reveal

Prompt: Reference: [product box photo], [neutral table surface photo]. Hands pull open a matte white product box on a light wood table, revealing a glass serum bottle inside. Natural daylight from left side. Slight camera wobble. 9:16 vertical.

The box flap opened with realistic paper resistance and the serum bottle read clearly on reveal. Lighting matched the reference window direction without any prompting on falloff. Generation: 2 min 08 sec.

2. Product-in-use

Prompt: Reference: [face moisturizer tube photo], [bathroom counter photo]. A person's hands squeeze a small amount of white cream from a tube onto fingertips, then press it into the back of their hand. Soft morning bathroom light. Shaky phone-camera feel. 9:16 vertical.

The tube label stayed legible through the squeeze and the cream had a translucent look that matched the reference. Hand-to-skin press motion was smooth without feeling artificial. Generation: 2 min 04 sec.

3. Before-after demo

Prompt: Reference: [cleaning spray bottle photo], [grimy stovetop photo]. A hand sprays a cleaning product onto a greasy stovetop, wipes with a white cloth, and the surface becomes visibly clean. Slightly overexposed phone-camera exposure. 9:16 vertical.

The spray arc was physically accurate and the wipe revealed a clean surface in one smooth motion. Product bottle shape held from the reference throughout. Generation: 1 min 58 sec.

4. Lifestyle moment

Prompt: Reference: [canned sparkling water photo], [outdoor picnic blanket photo]. A hand reaches into a white cooler and pulls out a slim can. Opens it. The can is set against a bright outdoor background with blurred grass. Warm afternoon sun. Feels like an Instagram story. 9:16 vertical.

The pull-from-cooler motion had realistic resistance and the can's label color matched the reference exactly. Background grass swayed naturally without any ambient motion prompting. Generation: 2 min 15 sec.

5. Pour

Prompt: Reference: [olive oil bottle photo], [marble counter photo]. A hand tips an amber glass bottle and pours a thin stream of golden oil into a white ceramic bowl on a marble surface. Close crop, handheld feel, natural warm kitchen light. 9:16 vertical.

The oil stream had convincing viscosity, caught the light mid-pour, and pooled at the bowl's base with accurate surface behavior. This is where Seedance's physics pull ahead of Kling on close product shots. Generation: 2 min 12 sec.

6. Hand-only product handling

Prompt: Reference: [lip gloss tube photo], [light pink flat-lay background photo]. A single hand holds a small lip gloss tube, uncaps it, and rotates the applicator wand toward camera. No face visible. Ring light glow from above. Slight hand sway. 9:16 vertical.

The wand applicator detail held up on close inspection and the cap-removal motion was clean. Hand morphing was minimal because the motion was slow and deliberate. Fast hand motions on close-ups are where this model struggles.

7. Vertical phone perspective

Prompt: Reference: [supplement bottle photo], [kitchen counter with morning light photo]. A hand holds a supplement bottle slightly below eye level as if filming themselves on a phone. Label faces camera. Turns bottle to show back panel. Morning window light. 9:16 vertical.

Label text from the reference stayed legible as the bottle rotated. The downward angle read convincingly as a self-filmed video. Generation: 2 min 06 sec.

8. Mirror selfie style

Prompt: Reference: [athletic wear product photo], [gym locker room mirror photo]. A hand holds a phone to a gym mirror. The phone screen shows a cropped view of someone wearing athletic shorts. Camera pulls back to show more of the outfit. Fluorescent gym lighting. 9:16 vertical.

The mirror-plus-phone composition rendered cleanly. Fluorescent lighting cast came through from the environment reference. This format runs slower due to double-frame complexity: around 2 min 25 sec.

Common failures

Over-polished output. Without an authenticity cue, Seedance defaults to clean production values. The clip looks like a commercial. Buyers scroll past it.

Hand morphing on close-ups. Fast hand motion on extreme close-ups produces finger geometry that doesn't track. The pour and wipe formats work because the motion is slow. A quick grab on a tight crop will break. Wider framing or slower motion prompts fix it.

Product label drift. Without a product reference, the model invents label text and it drifts frame to frame. Always attach the product photo. This is the single biggest improvement from multi-reference conditioning.

Mismatched lighting between cuts. Use the same environment reference photo for every clip in a multi-cut ad. Different environment references produce different ambient light and the cuts won't edit together.

Step-by-step on 8frame

1. Open 8frame and start a new canvas. Drop your product photo into the primary reference slot.

2. Add a second reference for the environment (table surface, bathroom counter, outdoor setting). Without both references, the model guesses at context and label consistency drops.

3. Paste your prompt, under 80 words. Include an authenticity cue and end with "9:16 vertical."

4. Set output to 9:16, 1080p, 30fps. Don't generate at 16:9 and crop; vertical native rendering is better for this use case.

5. Run. Expect around 2 minutes. While it generates, queue a variant with a different action or lighting cue.

6. Review at 1x then 2x speed. Physics issues that look fine at normal speed show up at 2x. If motion holds at 2x, the clip is usable.

FAQ

How does Seedance 2.0 multi-reference work?

You attach two images before generating: one for the product, one for the environment. The model uses both to constrain what it generates, which is why label text and textures stay consistent instead of drifting. Single-reference gives you one anchor; multi-reference gives you two. That's the difference between a usable product shot and a generic one.

Best aspect ratio for Seedance UGC ads?

9:16 vertical, generated natively. Don't generate 16:9 and crop. Seedance's composition logic for vertical is better than a cropped landscape output, and UGC framing (hand entering from below, product centered) depends on it. Use 9:16 for Reels, TikTok, and Shorts. For square feed posts, generate at 1:1 separately rather than cropping.

How do I keep the avatar consistent across cuts?

For human subjects, use Higgsfield Soul 2.0 instead. For product consistency (the more common UGC case), keep the same product reference image across every clip. Changing reference photos between clips is the main cause of label drift and color inconsistency in a multi-cut ad.

---

For the full multi-model chain from reference image to finished cut, the how to make a UGC ad with AI guide covers where Seedance fits alongside reference generation and post-production upscaling.

Clone the UGC ad workflow template from 8frame workflows and run these prompts against your own product photos in one place.