Skip to main content
AI video model
ByteDance
Seedance

Seedance 2.5

Seedance 2.5 combines a 30-second boundary, higher media counts, and video-editing reference input in one strict scene contract. Automatic duration is not underestimated at the 5-second default; Kyeo preauthorizes against the 30-second maximum.

Scene
Text / first-last frames / references
Output
480p / 720p / 1080p
Duration
4–30 seconds / automatic
Credits
102–5,700
30-second overview
  • Supports 4–30 seconds plus automatic duration
  • Accepts up to 30 images, 10 videos, and 10 audio clips
  • Reference-video and output duration are priced together
  • Supports MP4 / MOV output and returning the final frame

Popular searches: How do I use Seedance 2.5? · How much does Seedance 2.5 cost? · How is Seedance 2.5 automatic duration priced?

Examples

Cinematic scene

Cinematic 15 second commercial fashion promo shot on 35mm film with warm natural light and sub bass audio beat. Eye level static shot of a studio floor littered with scattered magazine pages and sleek minimalist clothing racks in the warm toned background. A stylish subject sits in a clean white modern chair at the central focal point. Every two seconds, rapid whip pan motion blur transitions cut between subjects seamlessly changing into distinct street fashion tracksuits: a dark gray ribbed tank top, an oversized purple and black panel jacket, a black and white hooded track jacket, a light blue zip up sweater, a soft pink contrast hoodie, and a vibrant yellow three stripe streetwear outfit. Natural human motion with fluid, subtle body rhythm and expressive hand gestures synchronized perfectly to an upbeat Afrobeat track with deep sub-bass and crisp percussion. Realistic skin textures, natural fabric folds, micro shadows, and organic camera sensor grain. No text or logos. Photorealistic depth of field, warm interior ambient lighting, cinematic color grading.

Cinematic scene

Concept: "Rainy Afternoon Errand" Create an authentic 15-second analogue home-video recording filmed with a handheld Hi8/VHS-C camcorder. The camcorder is never visible. The audience only sees the recorded footage, as if watching a digitized family videotape. This is not a vlog, commercial, or cinematic production—just a friend casually filming another friend during an ordinary day. Use authentic analogue camcorder aesthetics: soft SD resolution, subtle VHS tape grain, faint interlacing, analogue colour bleed, mild tracking imperfections, slight autofocus hunting, natural handheld shake, tiny framing errors, realistic motion blur, washed-out colours, and faint tape hiss. Avoid HDR, beauty filters, cinematic grading, perfect stabilization, or modern digital sharpness. A young Korean woman in her early twenties wears a fitted pastel crop top, mid-rise vintage blue denim shorts, white ankle socks, classic white sneakers, and a lightweight oversized windbreaker tied around her waist. She carries a small canvas shoulder bag, and her appearance reflects a nostalgic late-90s/early-2000s Korean street style without referencing any specific year. The scene takes place on a quiet residential street after a light summer rain. The pavement is still damp, small puddles reflect nearby buildings, bicycles are parked outside shops, neighbours walk by with umbrellas, and local cafés and convenience stores create a peaceful everyday atmosphere. The footage begins as she exits a small neighbourhood convenience store carrying a drink and a paper shopping bag. She pauses under the awning, looks up at the cloudy sky, and smiles as a light drizzle begins again. She playfully extends her hand to feel a few raindrops before noticing the unseen camera operator. She laughs, gives a friendly wave, briefly twirls once in the drizzle, then continues walking down the wet street. The camera follows naturally behind as reflections shimmer on the pavement and a city bus quietly passes in the background. The recording ends naturally as she disappears farther down the street. Generation Requirements Produce one continuous 15-second recording with no cuts or transitions. Maintain consistent identity, realistic anatomy, natural walking, authentic human expressions, believable interactions, and physically accurate rain effects. Include realistic ambient sounds such as soft rain, distant traffic, birds, footsteps on wet pavement, and neighbourhood ambience. Avoid AI artifacts, duplicated pedestrians, warped anatomy, floating objects, modern cinematic effects, or anything that reveals the footage as AI-generated. The final result should feel indistinguishable from a genuine nostalgic analogue home video recorded by a friend.

Cinematic scene

A cinematic lifestyle fashion video featuring a beautiful young Asian woman with long flowing dark hair. Opening shot uses an ultra-wide low-angle lens as she reaches her hand toward the camera under a dramatic blue sky with streaked clouds, creating a dynamic perspective effect. Smooth transition to an urban bus stop where she leans against a transit map wall, wearing silver headphones and a dark striped sweater, gazing thoughtfully into the lens. Final scene inside a minimalist modern café with abstract wall art, where she sits elegantly at a white table wearing a light green knit top and vibrant orange skirt. Soft natural daylight, shallow depth of field, smooth camera movements, realistic skin texture, subtle hair motion, emotional storytelling, clean aesthetic, commercial fashion photography style, cinematic color grading, bokeh background, ultra-realistic, 4K, high detail, luxury lifestyle advertisement, professional videography, smooth transitions, masterpiece quality.

Cinematic scene

Create a realistic fashion GRWM featuring a trendy Korean woman wearing a black fitted crop top, relaxed blue jeans, white sneakers, and a small crossbody bag. The video opens with her waving at the camera while sitting at her vanity. She does her makeup, curls her hair, picks her outfit from the wardrobe, wears sunglasses, packs essentials into her handbag, takes a mirror selfie, and leaves for a luxury shopping district filled with boutiques and cafés. Premium lifestyle commercial, cinematic camera movement, soft daylight, photorealistic, 4K.

Cinematic scene

Create a premium ultra-cinematic fashion film with flawless realism and seamless visual consistency. Duration: 15s Aspect Ratio: 16:9 Frame Rate: 24fps Lens: 85mm prime with extremely subtle handheld breathing motion. Use the attached storyboard image as the exact visual reference for the main character. Maintain identical face, hairstyle, hair length, hair texture, hair volume, skin tone, facial proportions, body proportions, and outfit throughout the entire video. Absolute character consistency. Never redesign the character. Style: Japanese airy film aesthetic blended with luxurious Indonesian nightlife. Kodak Portra 400 cinematic color science, fine film grain, soft bloom, warm neon glow, documentary-style realism. Color Palette: Deep crimson red, warm golden lights, black marble, champagne gold, luxury interiors, vibrant neon lighting, elegant city night, natural skin tones. Lighting: Soft cinematic lighting with a warm rim light. Environment: Foreground: Softly blurred chandeliers, cocktail glasses, luxury decorations, colorful neon bokeh. Midground: Premium rooftop lounge, luxury nightclub, elegant elevator lobby, stylish dance floor. Background: Jakarta skyline, skyscrapers, luxury cars, sparkling city lights. Atmosphere: Glamorous, fashionable, energetic, modern Indonesia. Main Character: • Exceptionally beautiful 20-year-old Indonesian woman • Tall and elegant • Fair glowing skin • Long black hair with soft natural waves • Natural glamorous makeup • Elegant fitted red cocktail dress • Black high heels • Small black clutch bag • Silver earrings • Minimal necklace • Maintain the identical hairstyle throughout the video SHOT 1 (0–2s) Luxury apartment. She adjusts her wavy hair, applies lip gloss, smiles confidently in the mirror under warm ambient lighting. SHOT 2 (2–4s) She sprays perfume, picks up her black clutch, and walks gracefully toward the door as the camera follows behind. SHOT 3 (4–6s) Luxury elevator opens. She walks through the elegant lobby. A luxury sedan waits outside. The valet opens the door. She enters as Jakarta's city lights glow. SHOT 4 (6–8s) She arrives at a premium rooftop lounge. Warm neon lighting. Guests socialize while the handheld camera follows beside her. SHOT 5 (8–10s) She meets close friends. They toast with colorful mocktails and laugh naturally in a classy atmosphere. SHOT 6 (10–12s) She casually dances with relaxed movements. Her long wavy hair flows naturally beneath elegant moving lights. SHOT 7 (12–14s) She steps onto the rooftop terrace. A gentle breeze moves her hair. She admires the Jakarta skyline before smiling warmly at the camera. SHOT 8 (14–15s) The camera slowly pulls back. She waves goodbye as the skyline sparkles. Natural fade to black. Audio: Luxury lounge ambience, soft crowd chatter, footsteps, elevator doors, mocktail glasses clinking, gentle background dance music, night breeze. No narration. No dialogue. Negative Prompt: No body deformation, facial distortion, plastic skin, excessive beauty filters, CGI appearance, hairstyle changes, outfit changes, character redesign, character inconsistency, hard cuts, white flash transitions, watermark, logo, or subtitles. Ultra-realistic documentary-style cinematography with perfect character consistency.

Cinematic scene

15-second vertical 16:9 fashion UGC GRWM video. Super casual real smartphone home-video footage of a young woman getting ready for a spontaneous weekend night out. The video feels like an authentic memory captured on her phone and later posted as a GRWM. Natural, imperfect, intimate, stylish, and believable. No commercial polish. MAIN SUBJECT Young Korean woman in her early 20s, natural everyday appearance, realistic skin texture, minimal makeup, warm expressive personality, long dark wavy hair with loose natural strands. She wears a fitted white ribbed tank top and relaxed low-rise blue jeans at the beginning of the video. Final outfit: oversized black leather jacket over a fitted white top, dark charcoal mini skirt, sheer black tights, black knee-high boots, small black shoulder bag, and minimal silver jewelry. Maintain identical facial identity, skin tone, body proportions, hairstyle, and appearance throughout the entire video. Clothing and accessories must remain physically consistent once introduced. LOCATION Small lived-in bedroom with a full-length mirror, unmade bed, clothing casually placed on a chair, open wardrobe, small vanity, cosmetics, jewelry tray, handbag, shoes, and warm apartment lighting. Soft blue evening light enters through the window while a warm bedside lamp illuminates the room. The environment should feel personal and naturally messy rather than staged. No studio environment. No luxury showroom. No commercial set. VISUAL STYLE Super casual real smartphone home video footage. Authentic social-media GRWM aesthetic. Realistic smartphone dynamic range, natural skin texture, imperfect exposure, subtle digital sharpening, ordinary indoor noise, realistic motion blur, believable shadows, and natural color reproduction. The footage should feel spontaneously recorded at home by a real person. No beauty filters. No cinematic color grading. No artificial film grain. No CGI appearance. No overly perfect skin. No fashion-commercial polish. CAMERA STYLE Super casual real smartphone home video footage captured by the creator herself. Slight authentic handheld shake, imperfect framing, occasional awkward camera positioning, small arm movements, natural walking bounce, subtle autofocus hunting, realistic exposure adaptation between window light and warm room lighting, slight smartphone sharpening, normal frame rate, natural motion blur, and occasional focus shifts. The phone is frequently repositioned between her hand, the bed, and the mirror. Use casual selfie footage, handheld rear-camera footage, mirror recording, close-up phone footage, and spontaneous room-level framing. Some compositions should be slightly off-center or imperfect. No professional stabilization. No gimbal. No cinematic tracking. No dramatic lens changes. No artificial depth of field. No polished fashion-film cinematography. PERFORMANCE Natural creator behavior. She is not acting like a model. She casually talks to the phone, makes small spontaneous reactions, checks her outfit, fixes her hair, laughs at herself, and moves naturally around the bedroom. Dialogue: “Get ready with me for tonight. I had no idea what to wear, but I think this might actually work.” Natural conversational delivery with realistic pauses, breathing, blinking, and accurate lip synchronization. ⸻ STORYBOARD 00:00–00:03 — “WHAT DO I WEAR?” Selfie-style handheld phone footage. She sits on the edge of her bed surrounded by two or three clothing options. She looks into the phone and says: “Get ready with me for tonight.” She quickly holds up a jacket toward the camera, makes an unsure expression, then looks toward the clothes on the bed. The framing is slightly crooked as she adjusts the phone in her hand. She laughs quietly. 00:03–00:06 — OUTFIT CHANGE Quick natural handheld transition. She is now wearing the white top, dark charcoal mini skirt, sheer black tights, and oversized black leather jacket. She stands in front of the full-length mirror while holding the phone. She pulls the jacket into position, looks at herself, then looks at the phone screen. She says: “I think this might actually work.” The camera briefly loses focus as she moves closer to the mirror before recovering naturally. 00:06–00:09 — ACCESSORIES Close handheld phone shot from the vanity. She picks up small silver earrings from the jewelry tray. Cut naturally to a mirror angle as she puts them on. She quickly runs her fingers through her hair, allowing loose strands to fall naturally around her face. She grabs the small black shoulder bag from the bed. Natural jewelry movement and realistic hand interaction. 00:09–00:12 — SHOES + FINAL CHECK Low casual phone angle. She sits on the edge of the bed and pulls on her black knee-high boots. The phone is slightly tilted, creating an imperfect home-video composition. She stands up, grabs her bag, and looks at herself in the mirror. She turns slightly from side to side to check the outfit. Natural fabric movement, realistic leather reflections, believable body mechanics. 00:12–00:15 — FINAL GRWM MOMENT Handheld mirror shot. She steps backward and gives the phone a quick full-outfit view. She adjusts the shoulder bag and lightly fixes her hair. She looks directly into the phone camera, smiles naturally, and says: “Okay… I’m actually obsessed.” She gives a small laugh, picks up her keys, and walks toward the bedroom door. The phone moves slightly with her hand. Recording ends naturally mid-motion. ⸻ AUDIO Authentic smartphone microphone recording. Natural bedroom ambience, subtle room tone, clothing movement, jewelry sounds, footsteps, bag movement, keys lightly jingling, and realistic household sounds. Her voice should sound naturally recorded in the room rather than professionally recorded. Very subtle background music may be present at low volume, similar to casual social-media GRWM content, but natural voice and room ambience remain dominant. No cinematic sound design. No exaggerated transitions. PHYSICAL REALISM Realistic human anatomy and hand proportions. Natural walking, sitting, standing, turning, and dressing movements. Correct hand-to-clothing interactions. Realistic leather jacket movement. Natural skirt and fabric folds. Realistic tights and boot interaction. Natural hair physics with individual strands moving during head movement. Realistic jewelry reflections and movement. Correct mirror reflections. No duplicated objects. No warped fingers. No impossible reflections. No floating clothing. No unnatural body mechanics. CONTINUITY LOCK Maintain the exact same woman throughout the entire video. Preserve facial identity, skin texture, hairstyle, body proportions, and natural appearance. Preserve the bedroom layout and lighting direction. Once the final outfit appears, maintain the exact same jacket, top, skirt, tights, boots, jewelry, and handbag. No spontaneous wardrobe changes. No changing accessories. No hairstyle changes. GENERATION REQUIREMENTS 15 seconds. 9:16 vertical. Super casual real smartphone home-video aesthetic. Authentic GRWM behavior. Realistic handheld phone movement. Natural imperfect framing. Realistic autofocus hunting. Natural exposure adaptation. Normal frame-rate motion. Accurate lip synchronization. Consistent identity. Consistent wardrobe. Realistic hair, fabric, leather.

At a glance

Is this model right for you?

First/last-frame input and reference media are mutually exclusive, and a last frame cannot be submitted alone. Reference videos and audio clips must each be 2–30 seconds, with separate 30-second aggregate limits. True media formats and image/video geometry are verified before credits are reserved.

What it does best
A video generation and editing entry point designed for longer output and large sets of multimodal references.
Who it is for
A good fit for complex storyboard references, video editing, creative experiments up to 30 seconds, and continuous workflows that need the final frame returned.
Why use it on Kyeo AI
Automatic duration is preauthorized against a provable maximum output, while signed media-duration evidence enters pricing to prevent under-reservation.

Key facts

Quickly evaluate whether this model fits your business use case.

Reference images
Up to 30
Reference videos
Up to 10; ≤30 seconds total
Reference audio
Up to 10; ≤30 seconds total
Output formats
MP4 / MOV

Other names

The same model may have different names in official documentation, community discussions, and API integrations. This section brings them together for you.

Seedance 2.5 multimodal
Seedance 2.5
Seedance 2.5 video editing

Common questions

These practical questions focus on the task, input conditions, and result requirements you should confirm before choosing.

How do I use Seedance 2.5?
How much does Seedance 2.5 cost?
How is Seedance 2.5 automatic duration priced?
How much reference media can Seedance 2.5 accept?
How long can a Seedance 2.5 video be?

Decision guide

Still deciding? Start with these key decision points.

1
Choose 2.5 for longer output or more media

When a 15-second or 9 / 3 / 3 limit is too low, 2.5 provides a wider contract.

2
Budget for the maximum with automatic duration

When -1 lets the model choose duration, Kyeo conservatively reserves for 30 seconds, which raises the required credit balance.

3
Remove redundant references first

A higher limit does not mean every slot should be used. Fewer conflicting assets can reduce both cost and instruction ambiguity.

Popular comparisons

Compare common alternatives on the same task to clarify differences in inputs, controls, and cost.

How do the media limits of Seedance 2.5 and Seedance 2.0 Mini differ?
How do the higher-resolution options and prices of Seedance 2.5 and MiniMax H3 compare?

Model comparison table

Not sure which model to choose? Compare the key differences between this model and its alternatives at a glance.

Duration and media
Seedance 2.5
Up to 30 seconds; 30 images / 10 videos / 10 audio clips
Seedance 2.0 Mini
Up to 15 seconds; 9 images / 3 videos / 3 audio clips; cost-efficient generation
MiniMax H3
Up to 15 seconds; 9 images / 3 videos / 3 audio clips; native stereo sound
Resolution and current pricing
Seedance 2.5
480p / 720p / 1080p; 28 / 63 / 158 credits per output second without reference video, or 17 / 38 / 95 per input-plus-output second with video
Seedance 2.0 Mini
Up to 720p; 3.8 / 8.2 credits per second without video, or 2.4 / 5 with video
MiniMax H3
Up to 2K; 8 / 13 credits per input-plus-output second at 768P / 2K

Real-world experience

Practical guidance based on public sources, current on-site limits, and representative tasks.

Automatic duration cannot be priced from guesswork

Preauthorizing against the contract maximum is the only way to guarantee that the task is not under-reserved before creation.

Video editing includes input time in the cost

Reference-video duration is added to output duration, so long input clips can substantially increase preauthorization.

Capabilities

30-second output boundary

Choose an integer duration from 4–30 seconds or let the model decide automatically.

High-count multimodal references

Combine images, videos, and audio clips in a reference scene.

Output workflow controls

Choose MP4 or MOV and optionally request the final frame.

Use cases

Long storyboard previsualization

Organize a longer shot from multiple motion clips and character images.

Video style editing

Submit a short reference video and describe the intended visual change.

Continuous shot handoff

Return the final frame to prepare a first-frame input for the next clip.

Prompt tips

Write actions by time segment

Describe shots in sequence across the 30-second window to avoid contradictory simultaneous events.

Number each reference's role

State which subject, motion, or mood each image, video, and audio clip should supply.

Automatic does not mean unlimited

Design both content and budget around the 30-second output ceiling.

Why choose it

Higher limits for both output and input duration.
Combines a large number of multimodal references.
Automatic duration uses conservative preauthorization.
Output-format and final-frame controls are explicit.

What to know first

1080p is the highest output tier currently available.
Automatic duration can require a high credit balance before creation.
Kyeo applies a conservative 95 MB limit to each video.
The content-check switch does not replace human review.

Adjustable settings

Quickly evaluate whether this model fits your business use case.

Duration
Optional
Parameter type: Integer
Default: 5 seconds
-1 for automatic, or 4–30 seconds
Resolution
Optional
Parameter type: Select
Default: 720p
480p
720p
1080p
Output format
Optional
Parameter type: Select
Default: MP4
MP4
MOV

Credit cost

102–5,700 credits

Without reference video, 480p, 720p, and 1080p cost 28, 63, and 158 credits per output second. With reference video, their rates are 17, 38, and 95 credits per second, multiplied by verified total input-video duration plus reserved output duration. Kyeo keeps the raw decimal total and rounds up only after multiplication; automatic duration is preauthorized against the 30-second output maximum, then settled against actual consumption.

Credit-saving tip

Before generating images or videos in bulk, run an A/B test with the same input on this model and similar alternatives. Confirm the result first to avoid wasting credits.

Frequently asked questions

Related models

Still deciding? Take a quick look at these similar alternatives.

Seedance 2.0 Mini

Seedance 2.0 Mini combines text, first/last-frame, and multimodal-reference generation in one entry point. Reference video enters the pre-task pricing formula, so both media duration and output duration must be verified before task creation.

AI video model
15–150 credits

MiniMax H3

MiniMax H3 offers 768P and 2K output with text, first/last-frame, and image/video/audio reference scenes. Reference-video duration and every input image from image 6 onward affect pre-task authorization.

AI video model
32–406 credits

Wan 2.7 Video

Wan 2.7 supports text-to-video, first-frame or first-and-last-frame image-to-video, and video editing. It currently offers 720p and 1080p, so compare media requirements, duration, framing, and estimated credits together.

AI video model
16 credits/second at 720p; 24 credits/second at 1080p

Sources

This page is based on vendor materials, model-family documentation, and Kyeo AI integration settings.

Source note

Explore Seedance 2.5 in 480p, 720p, or 1080p: 4–30 seconds or automatic output, up to 30 images, 10 videos, 10 audio clips, and 102–5,700 credits. Without reference video, 480p, 720p, and 1080p cost 28, 63, and 158 credits per output second. With reference video, their rates are 17, 38, and 95 credits per second, multiplied by verified total input-video duration plus reserved output duration. Kyeo keeps the raw decimal total and rounds up only after multiplication; automatic duration is preauthorized against the 30-second output maximum, then settled against actual consumption.

Last updated: 2026-10-06
Seedance 2.5 multimodal current interface notes
Integration

Used to check the input modes, visible controls, media limits, and availability of the version currently connected to this site. The page and workbench show the actual options.