Guide
This page collects Seedance 2–class fast-model outputs organized by creative category rather than by run order. The same underlying model handles very different anime genres—each places distinct demands on motion, expression, camera, and timing. By comparing categories side by side you can see what the model resolves well within each genre and where prompt structure matters most. All clips are locally hosted third-party examples; this is not an official product showcase.
Source basis and reading boundary
These guides are written as third-party reference summaries, not official product documentation or support content.
The transformation category asks the model to execute a two-beat visual arc: a calm opening (school uniform, blue sky, eye close-up) then a dramatic switch (golden magic circle, goddess armor, moon staff, particle effects). The core challenge is transition timing—how sharply the model cuts between states, how long the eye shot holds, and how heavily the effects land. Eight runs of this genre are documented in the iteration-notes guide linked below; here they serve as one side of the category comparison.
Heavy VFX layering (magic circles, light pillars, particle trails), costume swap mid-clip, strong emotional shift (calm → determined), and subtitle rendering in Japanese. The model must balance spectacle with readability—too many effects at once can bury the character. Prompt tweaks like switching '邪笑' (wicked smile) to '微笑' (gentle smile) shift the entire emotional register of the goddess reveal.
The romance category flips the challenge: instead of VFX fireworks, the model must hold a single consistent character across a multi-beat emotional progression in one continuous take. The camera is locked to first-person POV (only a hand enters frame), and the character must cycle through relaxed → surprised → shy → flustered → pouting. Dialogue lines ('哎呀，别闹啦') and props (the ice cream) must persist throughout. This is closer to a dating-sim or visual-novel short than an action sequence.
Character-sheet-level identity lock (hair color, eye color, outfit details, accessories), naturalistic micro-expressions across an emotional arc, correct hand interaction from an off-screen POV character, prop persistence (ice cream visible throughout), spoken dialogue with matching lip performance, and 9:16 vertical framing. The prompt is much longer and more prescriptive than the transformation prompts because every beat is an interaction rather than a visual effect.
Transformation prompts are effect-heavy but structurally simple (two states + a switch). Romance prompts are effect-light but interaction-dense (one character, many emotion beats, prop + hand choreography). The model handles both within a single fast pass, but the failure modes differ: transformation clips risk burying the character under particles; romance clips risk breaking character consistency or dropping the prop mid-scene. Understanding which category your project falls into helps you write tighter prompts and set realistic expectations for how many re-rolls you will need.
First-person hand-only interaction with a character-locked anime girl. The prompt specifies a detailed character design (ash-blonde bob, violet eyes, gothic-nun dress), an emotional arc across multiple beats (head pat, cheek pinch, shy reaction), and 9:16 vertical framing. This category tests Seedance 2 Fast's ability to maintain character identity, hand anatomy, and multi-beat emotional continuity in a single shot.
Transformation sequences where a character shifts from everyday attire into a supernatural battle form. The dark magical-girl prompt places the character on a dusk rooftop surrounded by terrified classmates, triggering a black-red gothic-lolita transformation. The vampire-noble prompt unfolds inside a gothic demon-king palace, with bat interaction and a scythe summon. Both test costume morphing, particle effects, and dramatic staging.
Three takes of high-energy anime combat. The prompt for these runs was not recorded, but the outputs show fast-paced action with dynamic camera, impact effects, and character motion. Comparing three stochastic rolls of the same (unknown) prompt shows variation in composition and choreography.
Quieter categories that test the model outside of action and transformation. Daily-life slice-of-life captures an everyday moment; kawaii character showcase focuses on cute design and gentle animation; cartoon-style animation applies a Western cartoon aesthetic. Prompts for these four videos were not recorded.
School-uniform eye close-up → Greek goddess with golden magic circle. Two-part prompt, wicked-smile variant. See the iteration-notes guide for all 8 runs.
日式动漫，高中校服少女，眼神特写，藍天白雲的背景电影质感，高清流畅 日语字幕： 是時候開始了。 日式动漫，少女变身希臘女神，金色魔法阵，月亮法杖，优雅邪笑，动态特效 日语字幕：月亮賜我力量
Single consolidated cinematic prompt with full art direction, Japanese subtitle lines, and detailed costume + effect descriptions. Highest information density of the eight runs.
高品質日式動漫風，電影級分鏡感，10秒短視頻，高清細膩，流暢動畫，強烈變身張力與史詩戰爭氛圍。藍天白雲下，一名身穿日式高中校服的少女站立，微風吹動頭髮與裙擺。鏡頭以眼神特寫開始，少女由平靜轉為堅定，日語字幕：「さあ、始める時だ。」隨即地面展開巨大的金色希臘風魔法陣……（see iteration-notes page for full prompt）
9:16 vertical, first-person hand-only POV. A character-locked anime girl (ash-blonde bob, violet eyes, gothic-nun dress with purple bow) eats ice cream, gets her head patted, reacts with shy surprise, gets her cheek pinched, blushes and pushes the hand away with a flustered pout. Multi-beat emotional arc in one continuous shot.
女主角色為女主角設定為圖中這位日系二次元少女：身形纖細嬌小，氣質文靜、乖巧、柔弱可愛，帶有明顯的害羞感與洋娃娃般的精緻氣質。她有一頭淺灰金色、偏亞麻白金色的及肩短髮，髮絲柔軟輕盈，髮尾微微外翹，額前是柔軟整齊的瀏海，頭頂有一撮標誌性的上翹呆毛，瀏海上方中央佩戴一個深紫色蝴蝶結。她的臉型小巧，皮膚白皙，眼睛大而圓，瞳色是清澈明亮的紫羅蘭色，整體眼神帶有無辜、安靜與淡淡羞澀感。她穿著深紫灰色的哥德修女風短洋裝，胸前有大面積白色翻領、深色領帶狀裝飾、十字架飾品與整齊的小圓鈕扣，袖口有白色荷葉邊；下身是多層次短裙，裙擺下露出白色內襯，搭配淡紫灰色半透明褲襪與黑色瑪麗珍鞋。整體視覺像一位清冷柔軟、害羞內向、非常惹人憐愛的二次元少女。以這角色為基礎，保持人物外觀、髮型、服裝、臉部特徵與整體氣質完全一致，生成一支 10 秒、9:16 直式、模型 2.0 的高品質短視頻。畫面採用第一人稱主視角，主角本人不出鏡，全程只露出一隻手，不出現主視角的臉、身體或其他部位。整體氛圍偏日系戀愛感，甜美、曖昧、輕柔，突出女生嬌俏、害羞、讓人心動的感覺。開場時女主正自然地吃著雪糕，神情放鬆可愛，鏡頭近距離呈現她的表情與上半身。主視角的手忽然伸出，輕輕摸向她的頭頂，像是親暱地逗她。女主微微一愣，帶著嬌羞可愛、很軟很甜的少女聲音說：「哎呀，别闹啦，我还在吃雪糕呢。」之後女生轉過頭看向主視角，眼神中帶著害羞、小小埋怨與戀愛感。主視角的手再輕輕捏一下她的臉頰，動作自然短促。女生瞬間臉紅，耳根發熱，露出明顯的羞赧表情，眼神有些躲閃。接著女生立刻用左手把主視角的手輕輕甩開，不是生氣打人，而是害羞得受不了的反應。她紅著臉，用很害羞、很可愛、像對喜歡的人撒嬌抱怨的聲音說：「讨厌啦……都说了别摸我好吗？」最後女生用手輕輕摸著剛剛被捏過的臉頰，低頭或微微別開視線，臉上仍帶著紅暈與羞意，形成甜甜的戀愛感收尾。要求表情細膩自然，手部動作正常，臉部不變形，雪糕全程可見，角色身份一致，不要誇張搞笑，不要暴力，不要出現第二個人，只保留第一人稱手部互動。
9:16 vertical, 464×832. A long-haired schoolgirl in sailor uniform walks through a terrified crowd on a dusk rooftop, slams an ornate staff on the ground triggering black-red smoke and dark particles, then transforms into a golden-haired, red-eyed villain in a black-red gothic lolita battle dress. Japanese dialogue line: 「主の御名のもとに、滅びなさい。」 Dense VFX with dark-magic circles, wind pressure, and an intimidating aura.
日系漫画动画风格，高质量二次元动画感。场景为黄昏时分的学校天台，夕阳将天空染成橙红色，风吹过栏杆与制服，周围站着一群神色惊恐的同学，气氛压抑而紧张。 主角是一位长发有刘海的少女，起初穿着水手服校服，外表清秀却带着冷静危险的气质。她从惊恐的人群中强势地走出，伸手推开挡在前方的同学，步伐从容，表情高傲而冷淡。随后她从背后缓缓拿出一把精美华丽的法杖，将法杖重重敲击在地面上。 法杖敲击地面的瞬间，周围爆发出黑红色的烟雾、暗红粒子与不祥魔力波纹，黑红魔力迅速环绕她的全身，风压吹起她的长发与裙摆。少女以高傲、邪恶而优雅的语气用日语说道："主の御名のもとに、滅びなさい。" 随后她开始华丽变身。黑红色烟雾包裹全身，长发飞扬并逐渐变成金色长发，双眼化为鲜红色，身上的水手服逐渐变化为一套黑红色的暗黑洛丽塔战斗装，整体风格优雅、邪恶、带有恶魔贵族与反派大小姐气质。服装带有黑色蕾丝、暗红色缎带、层叠裙摆、精致手套、恶魔风装饰和高贵挑衅感。 变身完成后，少女优雅地站在黄昏天台中央，金色长发在风中飞舞，红色眼睛闪耀，身穿黑红邪恶洛丽塔战斗装，手持华丽法杖，露出轻蔑又挑衅的微笑。她轻轻抬手施展魔法，黑红色魔法球在法杖前端凝聚并旋转扩散，声音邪恶、优雅、带着绝对压迫感。 整体要求：高质量日漫动画风，黑暗魔法少女，优雅反派大小姐风格，动作流畅，镜头感强，变身华丽，魔法特效精致，黄昏学校天台氛围强烈。
Anime-style 2D. A white-haired, red-eyed noble girl in a black nightgown walks through a lavish gothic demon-king palace under moonlight. A bat lands on her hand; she speaks in Japanese: 「もういいわ、あなたたちでは無理よ。私が出る。」 Black-red magic erupts — roses, crimson particles, vampire crests — transforming the nightgown into a layered vampire-noble gothic lolita gown with a jeweled scythe. Closing line: 「さあ、ゲームを始めましょう。」
二次元日漫画风，高质量动画分镜感。场景是华丽的魔王城堡宫殿内部，哥特式设计，暗红、黑金与月光交织的奢华氛围，巨大的落地窗外是夜晚天空。一个白色长发、红色眼眸、容貌精致、气质高傲优雅的少女缓缓走来，身穿黑色睡裙风睡衣，步伐从容，神情冷淡，带着上位者的压迫感。 一只蝙蝠从窗外飞入，轻轻落在她的手上。少女垂眸看着蝙蝠，优雅而冷淡地用日语说道："もういいわ、あなたたちでは無理よ。私が出る。" 蝙蝠飞走后，少女轻轻挥手，周围爆发出华丽的黑红色魔法光效，伴随黑雾、暗红粒子、绯红光带、玫瑰花瓣和血族纹章，展开一段精致的变身演出。她身上的黑色睡衣逐渐变化成一套黑红色血族风贵族洛丽塔长裙，服装华丽精致，带有吸血鬼贵族元素、暗红玫瑰装饰、黑色蕾丝、宝石胸饰、层叠裙摆与精致哥特头饰，整体高贵、危险、优雅。 随后她抬起手，黑红色魔法在掌心凝聚，化作一把巨大而华丽的镰刀武器，镰刀带有暗红宝石、玫瑰装饰与血族花纹。变身完成后，她优雅地站在宫殿中央，白色长发与裙摆轻轻飘动，手持镰刀，露出从容又危险的微笑，最后用日语说道："さあ、ゲームを始めましょう。" 整体风格要求：日漫高质量动画风，角色美型，暗黑哥特，血族贵族少女感，优雅强大，特效华丽精致，动作流畅，镜头感强，气氛高贵而压迫。
High-energy anime combat with dynamic camera, impact effects, and character motion. First of three stochastic rolls.
Second roll of the same battle prompt. Compare composition and choreography differences from stochastic variation.
Third roll. Three takes of the same prompt show how much output can vary in action sequencing and framing.
A quiet everyday anime moment. Tests the model on low-action, atmosphere-driven content.
Cute character design with gentle animation. Focuses on appealing design and soft motion rather than action.
Western cartoon aesthetic applied through Seedance 2 Fast. Shows the model can go beyond anime into broader animation styles.
No. These are editorial groupings we use to organize third-party test outputs. The model itself does not expose genre categories.
Romance POV prompts are typically much longer because they must lock character identity, choreograph multi-beat interactions, specify camera constraints, and script dialogue. Transformation prompts can be shorter because the visual arc is simpler (state A → effect → state B).
In principle yes—a transformation that also has romance dialogue—but the more demands you stack, the more likely something breaks in a single fast pass. Most creators separate concerns and composite in post.
Documented on the dedicated iteration-notes page, which pairs each of the 8 clips with its exact prompt and explains what changed between runs.
Four category videos (ACG battle ×3, daily life, kawaii, cartoon) were generated during early testing before prompt logging was in place. The outputs are still useful as style references.
Most are generated at standard Seedance 2 Fast resolution. Transformation and romance clips are typically 10 seconds in 9:16 vertical format. Battle and showcase clips vary.
Describe the before and after states clearly: starting outfit, transformation trigger (staff slam, magic circle, bat), and final form with specific costume details. Include particle effects, color palette, and emotional tone. See the dark magical-girl and vampire-noble prompts above for working examples.
Guide
Field notes from eight consecutive generations of a Japanese anime school-uniform-to-Greek-goddess transformation: matching prompts to clips, what changes between runs, and what fast re-generation is useful for.
Open guideGuide
Write better Seedance 2.0 prompts: subject + action + camera + style formulas, @ reference tags, and practical before/after tips for text-to-video and image-to-video workflows.
Open guideGuide
Best practices for Seedance 2.0: prompt formulas, reference assets, camera and motion wording, and quality checks. Based on public guides.
Open guideGuide
Seedance 2.0 use cases: e-commerce ads, TVC, product demos, film previz, MV, education, real estate, and short narrative. Based on official blog and third-party case studies.
Open guide
