Seedance 2.0 Text-to-Video Guide: Prompts That Work

Seedance 2.0 text-to-video is the right workflow when you do not already have an image, product frame, or character reference. You write the scene from scratch and let the model create the visuals, motion, and audio.

The tradeoff is control. Text-to-video gives you freedom, but less identity stability than Seedance image-to-video or reference-to-video. The prompt needs to define motion clearly without trying to micromanage every pixel.

The biggest mistake is treating text-to-video like a search box. "A beautiful cinematic ad for a smartwatch" sounds clear to a human, but it does not tell Seedance what happens on screen. A better prompt tells the model what a camera operator would shoot: the watch sits on black glass, the second hand starts moving, cyan light catches the bezel, the camera slowly orbits at table height.

That is the difference between an idea and a shot.

Start With A Shot, Not A Mood

Before you write a Seedance text prompt, answer five questions:

Once those answers are clear, the prompt almost writes itself. You can add style later, but the scene needs bones first.

What Is The Best Prompt Formula?

Use a single-shot formula for simple clips and a shot-list formula for sequences.

[Subject] [does one action] in [specific setting], [camera movement], [lighting or atmosphere], [sound cue].

Example:

A young chef slides a copper pan across a stainless counter, low side-tracking camera, warm kitchen light, pan scrape and soft sizzle.

For 8-15 second clips, use labeled shots:

Shot 1: A cyclist waits at a red light in light rain, close side profile, traffic hum.
Shot 2: The light turns green and the cyclist pushes off, low tracking shot beside the wheel.
Shot 3: Wide shot as the bike crosses a reflective street, rain and tire sound.

Why this works: each shot is a camera instruction, not just a sentence. Seedance gets a starting frame, a motion beat, and a payoff. If you wrote the same idea as one long paragraph, the model would have to decide where the cuts go.

How Should You Pace A Clip?

Pacing depends on duration. A 4-second clip should be one action. A 15-second clip can support a setup, movement, and payoff.

If the output feels chaotic, reduce either the number of shots or the number of actions per shot.

For text-to-video, duration is not just a quality setting. It is a storytelling budget. Four seconds can carry one action. Fifteen seconds can carry a mini-story. If you ask for a full ad, a costume change, a camera orbit, a product reveal, and a reaction shot in 5 seconds, the model will compress everything until the clip feels confused.

Camera Language That Works

Seedance responds better to physical camera directions than to generic style words.

Good camera moves:

• slow push-in
• locked-off wide
• low side-tracking shot
• handheld follow
• clockwise orbit
• top-down macro
• crane pullback

Use one main camera move per shot. "Orbit, dolly, zoom, handheld, drone" in one shot usually produces confused motion.

Camera language is also how you control emotion. A locked-off wide shot feels observational. A handheld follow shot feels immediate. A slow push-in feels intimate. A low-angle dolly feels powerful. These words do more work than "cinematic" because they tell Seedance how the viewer is physically moving through the scene.

Sound Cues For Text-To-Video

Seedance can generate synchronized audio, so sound cues are not decoration. They help time events. Use sounds tied to visible actions.

Avoid "great music" or "emotional sound." If you need music timing, describe tempo, beat, or transition point.

Sound is where Seedance prompts can feel unusually alive. A lot of video prompts look good but feel silent. Even if you plan to replace the audio later, a sound cue can help the model understand timing: the cap click happens when the bottle opens, the splash happens when the shoe lands, the room tone drops when the reveal happens.

Prompt Examples By Aspect Ratio

16:9 Landscape

16:9. A silver train crosses a mountain bridge at sunrise, wide locked-off landscape frame, mist moving below, distant rail sound.

What it does: this is a calm establishing shot. The train gives horizontal motion, the mist gives atmospheric motion, and the locked frame keeps the landscape readable.

9:16 Vertical

9:16 vertical. A creator opens a package at a desk, camera close to hands, tape tear and cardboard tap, bright natural window light.

What it does: this is a creator-style product beat. The hands fill the frame, the sound is tied to a visible action, and the vertical crop is not fighting a wide composition.

1:1 Square

1:1 square. A glass perfume bottle rotates on black marble, top-down light sweep, subtle music-box chime, clean luxury product frame.

What it does: square product shots work best when the subject is centered and the movement is symmetrical. This prompt is built for feed posts and product thumbnails.

21:9 Ultrawide

21:9 ultrawide. A lone car drives along a coastal highway at blue hour, drone pullback, ocean wind and engine hum.

What it does: ultrawide is for space. The prompt uses a small subject inside a large environment, so the aspect ratio feels intentional instead of empty.

When Should You Switch To References?

Switch from text-to-video to references when the output must preserve a specific face, product, room, logo, outfit, or camera rhythm. Seedance 2.0 reference-to-video can use images, video clips, and audio clips, but the prompt needs to assign each file one role.

Use text-to-video for exploration:

• mood tests
• concept boards
• generic social clips
• landscape and b-roll ideas
• rough ad beats

Use image/reference workflows for production:

• exact products
• repeat characters
• branded objects
• visual continuity
• audio-synced edits

If you keep rerolling a text prompt because the product changes shape or the character looks different each time, that is not a prompting failure. It is a workflow signal. Move to image-to-video or reference-to-video and give Seedance the visual anchor it needs.

A Better First Draft Workflow

Use this when you are exploring from text:

1. Write the simplest possible one-shot version.
2. Generate it short, usually 4-6 seconds.
3. Keep the useful motion and remove the rest.
4. Add shot labels only after the single-shot version works.
5. Move to Standard quality only after the prompt is stable.

The goal of the first draft is not to create the final clip. It is to learn what Seedance understands from your wording. Once the motion language is working, then you can add polish.

Related Guides

• Seedance 2.0 prompt guide
• Best Seedance 2.0 prompts
• Seedance 2.0 image-to-video
• Best text-to-video AI generators