Try Seedance 2.5 with a focused text or reference-guided video brief. Check input mode, supported settings, and credit cost, then review the generated motion.

Start with one subject and action

Write a short scene with one subject, one action, and one camera direction. For example: A ceramic cup sits beside a window as steam rises and the camera slowly moves closer. Keep the first request focused enough to review.

Open Seedance 2.5 in the workflow that matches your starting material. Check the duration and output controls shown in the editor rather than assuming every mode shares the same inputs. Review the estimate before submitting.

Choose text, still images, or footage first

Use Text to Video for an imagined scene, Image to Video for a still-image anchor, or Video to Video when compatible footage should guide motion. Select the workflow before adding references.

Upload only the media needed for the brief. A product photo can guide appearance; a motion clip can guide an action when supported. The editor determines accepted input types, counts, and limits for the selected mode.

Prompt: location, subject, one action, and one camera direction.
Image reference: the silhouette, material, palette, or appearance that should stay recognizable.
Video reference: a specific rhythm, gesture, timing, or camera movement.

Assign each uploaded source a role

Name which input guides appearance, setting, or motion using the tags displayed in the editor. Describe the final shot and the details to retain. Do not ask conflicting references to define the same subject.

Review the generated face, product, and movement against their assigned sources. Reference support does not guarantee exact identity, choreography, or timing. If the result drifts, simplify one role or instruction before testing again.

@image1: identify the appearance, product, material, setting, or palette to preserve.
@video1: identify the movement, timing, performance, or camera rhythm to borrow as creative guidance rather than a frame-perfect copy.
@audio1: identify the uploaded audio direction; name its role instead of assuming that every sound should define the scene.

ImagineScenes Video to Video editor showing one uploaded image, video, and audio reference with @Image1, @Video1, and @Audio1 in the prompt beside the completed video result — A completed five-second Seedance 2.5 pass: the thumbnail badges show the assigned reference tags, and the prompt gives every tag one specific responsibility.

Set a concrete review target

Choose a supported length that lets you inspect the intended action without adding several unrelated events. Set framing for the destination and review the updated estimate after each control change.

Define a successful first test in observable terms: the cup remains recognizable, the steam moves gently, and there is no abrupt cut. A scene that misses one target gives you a specific instruction to refine.

Use a prompt with clear source responsibilities

Name the appearance source, action, setting, camera direction, and details to retain in separate clauses. The example below is a creative direction, not proof that the model will reproduce every requested feature.

EXAMPLE DIRECTIONUse @image1 for the frosted glass bottle, pale green liquid, and clean label area. Use @video1 only for a slow forward camera rhythm. Use @audio1 as the sound direction for a quiet outdoor morning atmosphere. Place the bottle on pale stone beside sliced citrus in early-morning light. Condensation forms and moves slowly on the glass while leaves shift in a light breeze. Keep the bottle centered, leave the right third clear for a headline, and do not add other products.

Keep the reference set small and compatible

Begin with the fewest inputs needed to define the shot. A single clear still may be enough for a product animation; additional media should supply a different missing role.

Check each file against the model's displayed limits. A valid upload confirms file compatibility, not that its face, logo, voice, or motion will be copied exactly.

If a result misses, adjust one reference or prompt clause and compare it with the previous version. Each new submission consumes credits; adding more sources is not an automatic fix.

Review the whole clip before another pass

Play the opening, middle, and closing moments. Check subject shape, hands, text, camera changes, and any sound before deciding whether the shot is useful. Download or save the version you want to compare.

Refine the visible issue with one change. Recheck the current controls and cost before another generation or a separate editing stage. A selected model name does not establish a universal quality or speed advantage.