Prompt Formats
2 Prompting Styles, 4 Formats
The way you prompt Seedance 2 UGC is very different from other models. There are two fundamental styles, each with their own format.
Structural Prompting
Be descriptive and explicit about exactly what you want. Define the subject, action, props, wardrobe, tone, and key visual details. You control the generation.
Freestyle Prompting
Describe the result you want, not the steps to get there. Focus on the outcome and let the model interpret. Less control, more creative freedom.
Format 1
Structural — Descriptive
A single paragraph that describes everything: who is in the frame, what they are wearing, what they are holding, how they move, and what they say. Written like a scene direction.
Color-Coded Example
Using @image1 as the product reference, create a realistic creator style review video featuring a woman wearing the LED mask while it is turned on. She is dressed in a soft robe and holding the remote controller in one hand. She walks slowly toward the camera while naturally moving her hands as she speaks, keeping the remote visible. Her expression feels genuine and pleased, like she is sharing a product she personally loves. She says, "This LED mask has been a game changer for me."
The 5 components that make a descriptive prompt structural
1
Tag the image reference
Use @image1 so the model knows exactly which visual asset to use.
2
Describe the scene clearly
Who is in the frame, what they are wearing, what they are holding, and what the product is doing.
3
Define the movement
How the subject behaves on camera: walking toward camera, moving hands naturally, keeping product visible.
4
Set the tone and expression
How they feel and look: genuine, excited, personal, like sharing a product they love.
5
Specify the dialogue
The exact words the subject should say, so the model matches mouth movement accordingly.
Format 2
Structural — Breakdown
Every important part of the generation is separated into its own clear instruction. Instead of hiding everything inside one paragraph, you break the prompt into controllable components.
Reference: @image1
Subject: woman wearing the LED mask
Product state: LED mask is on and glowing
Wardrobe: soft robe
Prop: remote controller in hand
Action: walking toward the camera
Hand movement: natural hand gestures while holding the controller
Dialogue: "This LED mask has been a game changer for me."
Tone: authentic skincare review
Style: realistic creator testimonial
Focus: keep the LED mask and remote clearly visible throughout
Best for
Complex products with specific handling requirements. When you need precise control over wardrobe, props, product state, and camera behavior.
When to use
Product labels need to be readable. When the product has text, logos, or specific visual features that must stay consistent throughout the video.
Format 3
Structural — Timestamp
Timestamps control not just what happens, but when it happens. You guide the model beat by beat, controlling pacing and the exact order of actions. This gives the best results for complex multi-action scenes.
Reference: @image1 the LED product
Subject: female creator
Setting: cozy nighttime bedroom or apartment room
Wardrobe: light fitted tank top, casual lounge shorts, minimal jewelry
Product: light therapy face mask with visible cable
Tone: casual, personal, skincare obsessed
[00:00-00:01] She casually sips from a glass, opening the video like part of her real nighttime routine.
[00:01-00:05] She grabs the light therapy mask, holds it up clearly toward the camera, and says, "My favorite thing, I am obsessed with this light therapy mask..." Her delivery feels natural, excited, and genuine.
[00:05-00:12] She turns the mask on so the red light becomes clearly visible, then puts it on her face and adjusts it naturally with her hands.
[00:12-00:13] She is now wearing the mask and casually scrolling on her phone, showing how it fits into her routine.
Why timestamps matter
In video prompting, it is not just about what happens, but when it happens. Timestamps help control the pacing, the order of actions, and how the scene unfolds. Instead of one general idea, you guide it beat by beat.
Format 4
Freestyle
Freestyle prompting is less about precisely structuring the input and more about describing the result you want. It focuses on the outcome rather than controlling every individual component.
"An influencer trying an LED mask and talking about the different settings, how she uses it every night, and how easy it is to control."
The goal is not to direct every variable. The goal is to communicate the result. This works best for quick ideation, A/B testing concepts, and products that do not require precise handling instructions.
Best for
Quick concepts and ideation. When you want to explore creative directions fast without writing detailed structural prompts. Great for A/B testing.
Watch out
The looser the prompt, the more room the model has to improvise. This is where strange product interactions, physics breaks, and random text appear. Use structural prompts when accuracy matters.
The Prompt Formula
[Image Reference @tag] → [Subject + Wardrobe + Product] → [Action + Movement + Camera] → [Tone + Style + Feeling] → [Dialogue / Spoken Line]
