Updated September 5, 2026.
Quick Answer
For Seedance 2.5 image-to-video, upload a clear starting image or product reference in a supported mode, select the output format, and describe one motion or camera change. State which visible details must stay fixed, then inspect a short result before adding a more complex action.
This guide covers source-image preparation, reference roles, a three-pass consistency test and three product-video prompt examples. Start from the Seedance 2.5 workspace entry and confirm its current input controls.
What Is Image-to-Video?
Image-to-video is an AI-video workflow that begins with a still image. The image gives the model a starting composition, visible product details, lighting direction, and visual identity. Your prompt then tells it what should move, how the camera should move, and what must remain unchanged.
For commercial work, this is usually more controllable than starting with text alone. A text-only prompt can create a beautiful but unfamiliar product. A prepared reference image gives the model a visual anchor: the exact bottle shape, packaging color, dashboard layout, character wardrobe, or hero frame you need to keep.
| Input | It should control | It should not be asked to control |
|---|---|---|
| Source image | Product design, opening composition, important colors, visible details | A complete multi-scene story |
| Motion prompt | Action, camera direction, pace, atmosphere | Every product detail from scratch |
| Continuity line | What must remain stable | A long list of unrelated effects |
| QA review | The next one change to make | Ten changes at once |
Choose the input role before uploading your photo
An image can serve different jobs. Check which role your chosen workspace exposes, because a first-frame input and a general visual reference express different intentions.
| Input role | What you are asking for | A useful starting instruction |
|---|---|---|
| Starting frame, if available | Animate from this composition | “Begin with the uploaded framing and keep the product centered.” |
| Product or character reference | Reuse appearance in a new composition | “Use this image for the mug design; place it on a plain desk.” |
| Style reference | Borrow lighting, material treatment or palette | “Use the soft side light and muted palette; keep the product design from the product image.” |
| Final frame, if available | Arrive at a planned end composition | “Finish on the supplied hero composition and hold it steady.” |
Match the source crop to the intended output. A tightly cropped square photo cannot reveal extra background in a vertical video without inventing it. Leave room around handles, hair, lids and other details that may approach the edge during motion.
A three-pass test for a single product photo
Keep the image, model, aspect ratio and duration fixed. These passes are a suggested experiment, not a promise of a particular result.
| Pass | Prompt change | Acceptance check |
|---|---|---|
| A: appearance baseline | Product and camera stay still; allow only a soft moving highlight | Does the silhouette and label area remain recognizable? |
| B: camera test | Replace the locked camera with one gentle push-in | Does moving closer introduce shape changes or cut off edges? |
| C: action test | Keep the accepted camera instruction and add one small physical action | Does the action work without damaging the accepted appearance? |
If pass A fails, improve the photograph or reference assignment before adding motion. If A passes but B fails, reduce the movement or widen the source crop. If C fails, return to B and try a simpler action. Save every result so a stronger first frame does not hide a weaker ending.
For an app demo, keep exact screenshots as an editing overlay when text or navigation must remain faithful. Use generated footage for the surrounding scene or motion background. The app-screenshot demo guide explains that production split.
The Four-Part Image-to-Video Workflow
1. Start with a commercially usable still
Choose the image you would be happy to use as the first frame of a finished ad. It should have one obvious hero subject, adequate separation from the background, and enough empty space for natural motion. For a product image, make sure the label, cap, edges, and silhouette are already correct before you generate.
Avoid trying to repair a weak source image with a longer prompt. If the packaging is tiny, the subject is cropped awkwardly, or the background is visually noisy, fix the source first. Image-to-video is an animation workflow, not a reliable way to redesign the opening frame.
2. Define one motion objective
Before writing, complete this sentence: “The viewer should notice that the product is ___.” For example: rotating slowly, catching a moving highlight, emerging from mist, receiving a gentle push-in, or changing from a close detail shot to a clean hero shot.
One motion objective makes results easier to evaluate. A prompt that asks for a product rotation, splash, changing location, morph, fast zoom, new characters, and text overlay at the same time makes it impossible to identify why a result failed.
3. Add one camera instruction
Use one camera movement that supports the product action.
| Goal | Good starting camera direction |
|---|---|
| Show material detail | slow macro push-in |
| Reveal the whole product | smooth pull-back |
| Add premium energy | slow orbital move |
| Make a static UI feel active | restrained lateral slide |
| End on an editable frame | stable frontal hero shot |
Camera language does not need to be complicated. “Slow push-in toward the bottle; keep the label sharp” is more testable than stacking several lens and movement terms. For a fuller method, use the AI video prompt formula to separate subject, action, camera, lighting, style, and audio.
4. State what must not change
Continuity is a positive instruction, not just a negative prompt. Name the few details with business value: product shape, cap color, label layout, material, logo placement when you own the rights, camera direction, and opening composition.
Use plain language such as: “Keep the bottle proportions, cap, label placement, and purple glass material unchanged. Do not introduce readable text or additional products.”
How to Prepare a Source Image
Use this checklist before uploading.
- The hero subject is large enough to inspect at a glance.
- The first-frame composition is intentional.
- Important product details are visible and not blocked by reflections.
- The background leaves room for the requested movement.
- Any brand assets, people, or locations are owned or licensed for this use.
- The image has no tiny text that must remain perfectly readable in motion.
Why tiny text is a poor starting point
Models can preserve broad shapes better than small, exact typography. If a product label must stay legally accurate, make the label readable in the source image, keep the move restrained, and plan to add mandatory typography later in a conventional video editor. Do not claim that the generated clip is a perfect replacement for regulated pack-shot production.
A Seedance 2.5 Image-to-Video Prompt Template
Copy this structure, then replace the brackets.
Animate the uploaded product image into a [duration] product video.
Main action: [one visible action].
Camera: [one camera movement and speed].
Lighting: [one controlled lighting change or atmosphere].
Continuity: keep [product shape, material, label placement, cap, color] unchanged.
Ending: finish on [final framing] with clean space for text added later.
Avoid: extra products, unreadable lettering, sudden cuts, distorted geometry, and new logos.
The format matters because each line has a job. The image already provides the visual starting point; the prompt should supply only the movement plan and the constraints needed for review.
Three Product-Video Prompt Examples
1. Skincare bottle: premium macro reveal
Animate the uploaded skincare bottle image into a polished product-film shot.
Main action: small water droplets move slowly down the glass while a soft highlight travels across the bottle.
Camera: slow macro push-in, centered on the bottle.
Lighting: cool violet-blue studio light with controlled reflections.
Continuity: keep the bottle proportions, dropper cap, label placement, and glass color unchanged.
Ending: stable close hero shot with empty space on the left for text added in editing.
Avoid: extra bottles, readable new lettering, warped glass, hands, and abrupt motion.
2. Packaging: clean shelf-to-hero transition
Animate the uploaded package image into a minimal commercial video.
Main action: the package stays in place while a soft shadow recedes and the product becomes the visual focus.
Camera: gentle pull-back from a close product view to a stable centered hero frame.
Lighting: warm key light gradually becomes neutral studio light.
Continuity: preserve the package silhouette, color blocks, and all existing design elements.
Ending: clean frontal hero frame with room above for a headline added later.
Avoid: extra packaging, new logos, people, fast cuts, and a changed product shape.
3. App screen: restrained interface movement
Animate the uploaded app screen as a short product-demo background.
Main action: add a gentle ambient highlight around the screen while the displayed interface remains static.
Camera: restrained lateral slide; no dramatic zoom.
Lighting: retain the original clean UI color system.
Continuity: keep the layout, icons, and visible interface hierarchy stable.
Ending: hold on the unchanged screen composition for an editor to add an exact screenshot overlay and callout.
Avoid: invented menu items, unreadable UI text, random clicks, hands, and unrelated screens.
For more copy-ready directions across ads, social videos, stories, and image-to-video, see 20 Seedance 2.5 prompt examples.
QA Checklist: Review Before You Regenerate
Evaluate the generated clip against the same checklist every time.
| Check | Pass condition | If it fails |
|---|---|---|
| Product identity | Shape and key details match the source | Repeat the continuity line; simplify the action |
| Motion | One intended action is easy to see | Remove secondary effects |
| Camera | Movement supports the product | Replace multiple moves with one move |
| Readability | Hero object stays visible | Use a wider shot or slower pace |
| Ending | Final frame is useful in an edit | State the exact final framing |
| Brand safety | No unapproved logos, people, or claims | Remove them from the source and prompt |
Change one variable after each review. If the product drifted, fix continuity before changing lighting. If the camera is too fast, fix camera speed before adding a new effect. This simple loop creates a usable record of what improved the result.
Common Image-to-Video Problems
The product changes shape
Reduce the requested action, mention the two or three details that matter most, and avoid asking for a dramatic transformation. A product image should anchor product identity; it should not be asked to survive unrelated scene changes.
The camera move feels chaotic
Replace every camera term with a single instruction: “slow push-in,” “gentle pull-back,” or “stable hero shot.” Then regenerate before changing anything else.
The result looks polished but does not sell the product
The prompt probably optimized atmosphere rather than the product. Start again with a simple product action and a camera move that keeps the object large in frame. For a structured diagnosis, use Seedance 2.5 prompt troubleshooting.
The model invents text or an extra logo
Explicitly request no new text or logos, and do not rely on the generated frame for exact brand typography. Add final legal copy, pricing, and calls to action after export with your standard editing tools.
Final Checklist
- Start from the exact opening frame you want.
- Give the model one product action.
- Use one camera direction.
- Name the details that must remain fixed.
- Keep on-screen text for post-production.
- Review identity, motion, camera, readability, ending, and brand safety.
- Change one variable per regeneration.
FAQ
Is image-to-video better than text-to-video for product ads?
It is usually a better starting point when the product’s appearance and opening composition must match an existing asset. Use text-to-video when you need the model to invent a scene from scratch; use image-to-video when you need to animate an approved visual.
How long should an image-to-video prompt be?
Long enough to define the action, camera, continuity, and ending. Extra style adjectives are not a substitute for a clear motion plan. Begin with four to six lines and add detail only when a specific review failure requires it.
Should I use a product label as a reference?
Use the label in a source image if you own the asset, but do not assume small typography will remain perfectly legible in every generated frame. Add mandatory copy in post-production.
What is the best first camera move for a product image?
A slow push-in or gentle pull-back is usually the safest first test. Both are easy to judge and help keep the product visible.
Related Articles
- Seedance 2.5 image-to-video workflow — open the model workspace and apply this guide with an approved reference image.
- AI Video Prompt Formula for Seedance 2.5 — write clearer subject, action, camera, lighting, style, and audio instructions.
- 20 Seedance 2.5 Prompt Examples — adapt prompt structures for product ads, social clips, and explainers.
- Seedance 2.5 Prompt Troubleshooting — fix motion drift, product changes, camera jumps, and weak continuity.
- AI Video Camera Movements — choose a controlled push-in, pull-back, orbit, tracking shot, or final hero hold.
Put the guide into practice
Ready to test the workflow? Open the Seedance 2.5 image-to-video workspace, use one approved product image, and start with the template above. If you need to develop a scene without a starting image, try the text-to-video workflow.
