Seedance 2.5 Anime Prompts: 20 Templates for Elser AI Characters

A useful AI video prompt is not a bag of impressive adjectives. It is a compact production brief. It tells the model what must remain the same, what changes during the shot, how the camera observes that change, and what should not appear.

The templates below are designed for original characters created in Elser AI and animated with Seedance 2.5. Replace the bracketed fields with your own approved character and scene information. This is a creative workflow rather than an announcement of a formal product partnership. Availability may also vary: Dreamina's official Seedance 2.5 page was still labeled “Coming Soon” when verified on August 4, 2026.

The Prompt Structure Behind Every Template

Use six blocks:

1. Continuity: face, age, hair, outfit, prop, and rendering style to preserve.
2. Starting state: what exists in the opening reference.
3. Primary action: one visible change.
4. Performance: emotion, posture, and intensity.
5. Camera and environment: framing, movement, light, and secondary motion.
6. Exclusions: unwanted characters, cuts, text, redesigns, or effects.

Do not keep every bracket. Turn the completed template into natural instructions. Use only references you own or have permission to use, and avoid prompts that imitate a living artist or protected character.

Quiet Acting Prompts

1. The realization

Preserve [CHARACTER]'s approved adult face, [HAIR], [OUTFIT], [SIGNATURE DETAIL], body proportions, and clean [2D/3D] anime rendering. Starting from [OPENING FRAME], they read the note once, pause, then look back at one specific line as understanding slowly reaches their expression. Locked medium close-up with a very gentle push-in; one blink, subtle breathing, fingers tightening slightly on the paper. [LIGHT] remains stable. No tears, no dialogue, no costume change, no extra person, no text generated inside the scene.

2. Held-back laughter

Keep the same [CHARACTER] identity and outfit. They try to remain serious while listening off-screen, press their lips together, glance aside, and finally allow one small smile. Static close-up, restrained facial motion, shoulders relaxing at the end. Preserve eye shape and apparent age. No exaggerated comedy face, no cut, no added accessories.

3. Listening reaction

Preserve [CHARACTER] from the references. In [LOCATION], they listen to an unseen speaker at frame [LEFT/RIGHT], shift their gaze once, and lower the object in their hands as concern replaces confidence. Medium shot, slow lateral drift, soft environmental movement only. Maintain screen direction and consistent anime linework. No visible speaker, no lip movement, no text.

4. Waking from a dream

Same approved character, sleepwear, room design, and color palette. [CHARACTER] wakes gradually, focuses on the ceiling, then sits halfway up after noticing [OBJECT]. Dawn light moves softly through curtains. Overhead-to-medium camera transition kept smooth and slow. Natural breathing and cloth movement; no sudden horror distortion, no outfit change, no additional character.

Movement and Camera Prompts

5. Side-tracked walk

Preserve [CHARACTER]'s identity, proportions, [OUTFIT], and [PROP]. They walk at a steady pace through [LOCATION] while the camera tracks beside them in a medium-full profile shot. [HAIR/CLOTH] responds gently to [WIND/RAIN]. They notice [EVENT] and slow without stopping. Keep foot contact stable and background direction coherent. No cut, no running, no new character, no camera orbit.

6. Controlled reveal

Begin on a close view of [OBJECT] in [CHARACTER]'s hand. The camera slowly pulls back to reveal the same character standing in [ENVIRONMENT]; they lift their gaze toward [TARGET]. Preserve outfit, hand placement, object scale, and lighting direction. One continuous movement, cinematic but restrained. No object transformation, no title text, no extra limb.

7. Quarter orbit

Preserve the approved [CHARACTER] face, hairstyle, costume construction, and age. Starting from a three-quarter medium shot, the camera makes a slow 25-degree orbit toward profile as the character turns only their eyes toward camera. Use the turnaround reference for the newly visible side. Stable background parallax and consistent cel shading. No full rotation, no hairstyle change, no jewelry.

8. Entrance with purpose

In the approved [LOCATION] opening frame, [CHARACTER] opens the [DOOR/GATE], steps inside, and stops after seeing [OFF-SCREEN DISCOVERY]. Medium-wide camera gently tracks backward to preserve distance. Maintain exact screen direction, costume, and prop position. Environmental particles move subtly. No cut, no duplicate character, no generated signage.

Action Prompts

9. Sword draw without a fight

Preserve the original adult character and non-franchise costume. [CHARACTER] plants their feet, reaches with the right hand, and draws the approved sword in one clean controlled motion, ending in a defensive guard. Medium-full locked shot with a slight push-in after the blade clears the sheath. Keep hands attached correctly, sword length constant, and left/right orientation stable. No opponent, no attack, no sparks, no cut.

10. Short chase beat

Same [CHARACTER] identity and outfit. They run through [LOCATION] toward frame right, vault one low obstacle, recover, and continue. Side-tracking camera at matching speed; background moves consistently, coat and hair follow momentum. Use readable 2D action timing without smear-frame facial distortion. No location change, no pursuer visible, no extra obstacle, no text.

11. Magic with a defined effect

Preserve [CHARACTER], costume, and [MAGIC DEVICE]. They raise the left palm and create one small [COLOR] geometric light above it; the light rotates once and illuminates the face and nearby wall. Slow push from medium to medium close-up. Keep the effect separate from fingers and preserve eye color. No explosion, no costume transformation, no floating symbols resembling text.

12. Two-character handoff

Preserve [CHARACTER A] and [CHARACTER B] from their separate approved references. A stands frame left and passes [PROP] with the right hand; B stands frame right, receives it with both hands, and looks at A. Locked two-shot, minimal body movement, prop scale and design remain constant. Maintain each face and costume. No merged hands, no third character, no cut, no dialogue.

Environment-Led Prompts

13. Rainy station pause

Same [CHARACTER] identity and outfit under the station canopy. They watch a train pass beyond the platform, then step closer to the edge without crossing the safety line. Medium rear three-quarter shot, gradual dolly-in, rain and reflected light moving beyond the canopy. Use the turnaround for rear costume details. No readable station text, no additional foreground person, no camera shake.

14. Windy rooftop decision

Preserve [CHARACTER], [OUTFIT], and [PROP]. On the rooftop at sunset, they tighten their grip on the railing, look toward the distant tower, and make a quiet decision shown by posture becoming upright. Wide-to-medium slow push, hair and loose fabric reacting consistently to strong wind. Preserve skyline and sun direction. No flying debris near the face, no cut, no transformation.

15. Cozy interior loop

Keep [CHARACTER] seated at the approved desk. They write one line, pause to sip tea, and return the cup to the same coaster while a small lamp flickers gently. Static medium shot designed for a seamless calm loop; hand and cup contact remain clear. Maintain room layout and warm palette. No camera move, no added objects, no steam covering the face, no text legibility required.

Social and Commercial Prompts

16. Vertical character introduction

9:16 composition with [CHARACTER] centered inside the mobile safe area. Preserve their approved identity and outfit. They enter from lower frame, stop, lift [SIGNATURE PROP], and look directly toward camera with [EMOTION]. Gentle upward camera tilt, clean background motion, readable silhouette. Leave uncluttered space in the upper third for text added later. No generated captions, no logo, no costume change.

17. Product discovery

Preserve the original fictional character and the exact approved [PRODUCT] shape, packaging colors, and label-free hero surface. In [SETTING], the character picks up the product, rotates it only 30 degrees, notices one feature, and smiles naturally. Medium product-and-character shot, slow controlled dolly, accurate hand contact, stable lighting. No invented claims, no altered package geometry, no readable text generated in-camera.

18. Channel bumper

Same [CHARACTER] and palette. They perform one signature gesture toward an empty graphic area as [SIMPLE ORIGINAL MOTIF] forms behind them from light and particles. Five-second clean animation, fixed camera, strong beginning and ending poses, uncluttered 16:9 composition. No brand logo generated by the model; reserve negative space for the official title added in editing.

Story Prompts

19. Ten-second emotional turn

Preserve [CHARACTER]'s identity, costume, and restrained anime performance. 0–3 seconds: they wait beside [OBJECT], pretending to be calm. 3–6 seconds: [SMALL EVENT] occurs off-screen and they turn. 6–10 seconds: they recognize [DISCOVERY] and their tension releases into [EMOTION]. One medium shot with a subtle push-in, stable environment, no cut, no extra character, no text.

20. Thirty-second micro-story

Preserve [CHARACTER], [SECOND CHARACTER/PROP], approved costumes, and [LOCATION] across the full scene. 0–6 seconds: establish [GOAL]. 6–14 seconds: [OBSTACLE] interrupts the action. 14–22 seconds: the character tries [SIMPLE RESPONSE]. 22–28 seconds: [REVERSAL] changes the meaning of the scene. 28–30 seconds: hold a clean ending pose for editing. Use [CAMERA PLAN] with no more than [NUMBER] motivated framing changes. Maintain identity, screen direction, lighting, and anime rendering. No random subtitles, no unplanned character, no franchise imagery.

How to Adapt the Templates Without Breaking Them

Start with the shortest template matching your shot. Replace brackets, remove irrelevant lines, and read the prompt once as a director. Can every action be seen? Does one instruction contradict the opening image? Is the camera trying to do too much?

Pair prompts with clean Elser AI assets: canonical portrait, costume sheet, prop reference, and shot-specific opening frame. Register at Elser AI if you need a place to create the character and storyboard before testing motion. The useful registration journey is character → references → shot—not signup → random generation.

Keep a prompt log with the exact references and selected result. When a prompt works, preserve its continuity block and change only action or camera. When it fails, diagnose face, prop, motion, composition, or style separately.

FAQ

Should Seedance prompts be long?

They should be complete enough to resolve continuity, action, camera, and exclusions. A simple close-up may need four sentences; a timed micro-story may need more. Length itself is not quality.

Do negative prompts guarantee clean results?

No. Exclusions reduce ambiguity but cannot guarantee an outcome. Strong opening frames and compatible references remain essential.

Can I copy these prompts exactly?

Use them as structures. Replace every bracket and delete instructions that do not match your scene. Generic copying produces generic video.

Why avoid generated text in the shot?

Critical titles, claims, and dialogue need exact spelling and placement. Add them during editing where you can verify them.

Conclusion

The best Seedance 2.5 anime prompts describe time while references protect appearance. Begin with continuity, direct one visible action, choose a camera that supports it, and state only the exclusions that matter.

Use the twenty templates as production scaffolding, not magic spells. Create the approved character and opening frame in Elser AI, test a quiet performance before complex action, and save every successful combination. A good prompt becomes more valuable when your team can explain why it worked.