Back to blog

Seedance 2.5 Image to Video: Turn Any Photo into Motion

Step-by-step guide to Seedance 2.5 image-to-video on Seedove: upload a reference photo, keep the subject’s identity intact, and write motion prompts that work.

Sep 1, 2026Seedove TeamSeedove Team
Seedance 2.5 Image to Video: Turn Any Photo into Motion

Seedance 2.5 isn't just a text-to-video model — give it a reference image and it animates that exact subject: the same face, the same product, the same style. This is the workflow for product shots, character clips, and "bring my photo to life" videos. Here's exactly how to do it on Seedove, and how to write prompts that make image-to-video shine.

When to use image-to-video

Use a reference image whenever the subject must look a specific way:

  • Product videos — your actual product, not a hallucinated lookalike
  • Character / portrait clips — a real person's identity preserved
  • Brand visuals — your logo, packaging, or illustration set in motion
  • Artwork animation — sketches, paintings, and renders that come alive

If the subject can be generic ("a surfer at sunset"), plain text-to-video is simpler. If it must be your subject, attach the image.

Step by step on Seedove

  1. Open the generator and sign in.
  2. Attach your reference image — click the image button in the message box, or simply paste the picture straight into the chat. You'll see a thumbnail pinned above your message.
  3. Describe the motion you want, not the image content (the model can already see it): "She slowly turns toward the camera and smiles, soft window light, shallow depth of field."
  4. The agent drafts a script card. Check that it shows your reference image thumbnail — that confirms the generation will use it. Adjust the model, aspect ratio, duration, and resolution on the card.
  5. Hit Confirm & generate. Seedance 2.5 renders a 5s or 10s clip that starts from — and stays faithful to — your photo.

One image per generation; attaching a newer image replaces the reference for the next one.

Writing prompts for image-to-video

The mindset shift: with text-to-video you describe everything; with a reference image you describe what changes.

  • Do: "Camera slowly orbits the product, studio lighting, the label glints" — motion, camera, light, mood.
  • Do: state identity rules when a face matters: "Preserve the person's exact face and identity; change nothing about their features."
  • Don't: re-describe the subject's appearance in detail — it can fight the reference image.
  • Don't: stack too many actions. One primary motion + one camera move is the sweet spot for 5–10 seconds.

Three starter templates

Portrait come alive

The person in the reference image slowly looks up and smiles gently, subtle hair movement in a light breeze, soft natural light. Identity perfectly preserved. Slow cinematic push-in.

Product showcase

The product in the reference image rotates smoothly on a clean studio pedestal, soft reflections sweeping across its surface, premium commercial lighting, slow 360° orbit.

Drawing process / artwork animation

The sketch in the reference image gradually gains realistic shading and depth, graphite strokes appearing in time-lapse, paper texture consistent throughout, silent and hypnotic.

Aspect ratio and duration tips

  • Match the aspect ratio to where the video will live: 9:16 for TikTok/Reels/Shorts, 16:9 for YouTube and ads, 1:1 for feeds.
  • 5s covers a single clean motion; choose 10s when the motion needs a buildup (e.g. a full drawing reveal).
  • 720p is the right default; drop to 480p for quick drafts and save 1080p/4k for final renders.

FAQ

Does image-to-video change the face in my photo? Seedance 2.5 is currently the best Seedance model at identity preservation, and stating "preserve the exact face/identity" in your prompt helps further. Minor stylization can still happen in fast motion — generate twice and pick the better take.

Which models support reference images? Every Seedance model on Seedove — 2.5, 2.0, 2.0 Fast, 2.0 Mini and 1.5 Pro — accepts one reference image. We still recommend 2.5 for faces.

Does it cost extra? No. Image-to-video costs the same credits as text-to-video for the same model, duration, and resolution — e.g. Seedance 2.5 at 720p is 100 credits for 5s.

Can I upload more than one image? One image per generation. For multi-shot stories, generate separate clips and cut them together.

Ready to animate your first photo? Open the generator — new accounts get free credits, enough for several image-to-video clips.