Seedance 2.0 Image-to-Video Guide: Prompts and Settings

Seedance 2.0 image-to-video turns a still image into a moving clip. The key is restraint: do not use the prompt to redesign the image. Use the prompt to say what changes over time.

If the uploaded image is a portrait, the prompt should describe the head turn, blink, light shift, or camera move. If the image is a product shot, describe rotation, reflections, water droplets, hand interaction, or a clean dolly. If you need a full scene from text instead, use the Seedance text-to-video workflow.

The best way to think about it: your still image is the set. Your prompt is the director walking onto that set and saying, "Now move the camera here. Let the light shift. Let the fabric breathe. Do not change the product."

That last part matters. A lot of image-to-video prompts fail because they ask for a transformation when the job is actually animation. If you upload a good product render, you probably do not want Seedance to invent a new product. You want the same object to catch light, rotate, steam, splash, or move through one clean beat.

Before You Animate An Image

Do a quick source-image check before spending credits. Seedance can animate a weak image, but it cannot fully repair one while also preserving identity.

If the source image is doing too much already, write a quieter prompt. If the source image is simple, you can use the prompt to add atmosphere.

What Settings Work Best?

Use shorter durations for identity preservation and longer durations only when there is enough action to justify the time. Seedance 2.0 supports 4-15 second clips on Morphed, with standard output up to 1080p and Fast output up to 720p.

The most common mistake is trying to turn one source image into a full commercial. A still image can support a strong moving shot, but not ten unrelated scene changes.

The Three Best Image-To-Video Formats

Most reliable Seedance image animations fall into three formats: product orbit, portrait life, and environmental drift.

Start with these before trying more ambitious sequences. They are not boring; they are the formats closest to how real commercial b-roll is shot.

How Should The Prompt Be Written?

Write image-to-video prompts in this order: image role, action, camera, environment change, sound, constraints. The first sentence should tell Seedance what the image is responsible for.

Use the uploaded image as the exact product identity. The product stays the same shape and color. The camera slowly pushes in while soft studio light moves across the surface, subtle reflection on black glass, quiet room tone.

This is stronger than:

Make a cinematic beautiful commercial from this image, premium, dramatic, high quality.

The first prompt gives the model motion and constraints. The second prompt mostly gives taste words.

If you only remember one rule, make it this: never let the prompt fight the image. When the image is already specific, write less about appearance and more about movement.

Product Image Prompts

Product clips should protect geometry first. If the bottle, shoe, watch, or package changes shape, the clip is unusable.

Use the uploaded image as the exact bottle identity. The bottle remains the same shape, label, and cap. Slow clockwise orbit camera, water droplets sliding down the glass, soft bathroom light, faint water sound.

Use the uploaded shoe image as the exact product. The shoe does not change design. It lands once on wet pavement in slow motion, low side camera, splash and rubber squeak.

Use the uploaded packaging image as the exact box design. The camera glides from left to right across the front panel, light catching the embossed logo, no extra text, soft studio ambience.

For ecommerce, one clean movement usually beats an over-edited sequence. Use the best image-to-video AI tools guide if you want to compare Seedance against Kling, Veo, Runway, and Sora, or start from the image to video AI page for a model chooser and copy-ready motion prompts.

Director note: product clips usually look more expensive when the camera moves slowly and the light does the drama. Fast camera moves make products feel cheap unless the brand is intentionally high-energy.

Portrait And Character Prompts

Portrait animation fails when the prompt asks for too much expression, camera motion, and scene change at once. Keep the face stable and move one thing at a time.

The face is the most fragile part of a portrait animation. A slight turn can feel alive. A wide smile, sudden laugh, dramatic head whip, and camera orbit in the same 5-second prompt will often change the person's identity. Treat portraits like close-up acting direction, not action scenes.

Use the uploaded portrait as the exact identity. The subject turns slightly toward a window, blinks once, hair moves in a light breeze, soft natural light, quiet room tone.

Use the uploaded character image as the exact character design. The camera pushes in slowly while the character looks up from a book, candlelight flickers on the face, no outfit change.

Use the uploaded fashion image as the exact outfit and pose reference. The model shifts weight and turns one shoulder toward camera, fabric moves naturally, clean studio floor, no face change.

When Should You Add An End Frame?

Use an end frame when the start and finish compositions both matter. Product before/after shots, pose changes, doorway reveals, and logo endings benefit from end-frame control. Do not use an end frame for simple motion like steam, rain, blinking, or a small push-in.

End frames are especially useful for ads because the last frame is often the thumbnail, CTA moment, or product lockup. If the ending matters more than the motion, give Seedance the final composition instead of hoping it lands there.

Why Audio Cues Help Image Animation

Seedance can generate synchronized audio. Even when the visual is the priority, a short sound cue improves timing. Write physical sounds, not abstract moods.

Useful cues:

• rain on glass
• shoe splash
• fabric rustle
• soft camera shutter
• bottle cap click
• cafe room tone

Avoid vague cues like "epic sound" or "cinematic audio." They do not tell the model what event to sync with.

Small sounds are usually better than big sounds. A cap click, glass tap, shoe squeak, or page turn gives the model a physical event. That is much easier to sync than "inspiring music."

Standard vs Fast For Image-To-Video

Use Fast when you are still choosing direction. Use Standard when the clip is likely to ship.

If you are writing many variants, draft in Fast, then rerun the winning prompt in Standard at the final aspect ratio.

A Simple Workflow That Saves Credits

Use this workflow when the clip matters:

1. Draft a 4-6 second Fast version at 720p.
2. Keep the best prompt structure, not necessarily the exact wording.
3. Remove any extra style words that did not affect the result.
4. Rerun the winning version in Standard at the final aspect ratio.
5. If identity changes, make the motion smaller instead of adding more constraints.

Most bad image-to-video iteration goes the other direction: people respond to a weak output by adding more words. Sometimes the fix is fewer words and a clearer movement.

Related Guides

• Seedance 2.0 prompt guide
• Best Seedance 2.0 prompts
• Seedance 2.0 Fast vs Standard
• Seedance 2.0 vs Kling