Flova Anime Crossover Production

Anime Crossover Production

Mission

用于动画跨界短片：为原创动画角色、魔法变身、群像合体、屏幕破壁和现实/动画世界交互生成可控分镜、资产与视频提示词。

JarvisHub Execution Model

• 主 Agent 负责编排：读取画布事实、整理任务 brief、按阶段同步 TodoWrite 进度，并把已确认的文本成果写入画布；媒体生成、等待、拼接和评审交给具备相应能力的执行 agent 或当前可用工具。TodoWrite 不是画布写入前置条件。
• 图像、视频和拼接交给具备媒体能力的执行 agent：brief 给出稳定输出身份、用途、真实参考 URL 和关键约束；需要下游引用时必须等待真实 imageUrl / videoUrl，不能把提交态当完成态。
• 多模态验收交给 critic sub-agent：只读取真实媒体并评审，不生成、不补素材。
• 当前已知 canvas 工具集没有通用音频生成、音频驱动口型、末帧抽取或视频直改工具；旁白、BGM、字幕、末帧承接和精准 lip sync 只能作为后期合成计划或 blocked 项，除非本轮工具列表明确暴露对应能力。

When To Use

Use this skill for:

• 90s hand-painted cel anime feel,
• magical transformation sequence,
• magical-girl group battle,
• anime OP/ED-like segment,
• character breaks through the screen,
• viewer pulled into another world,
• original anime character enters reality.

Do not use it to reproduce named protected anime characters, official costumes, logos, theme songs, or exact transformation shots. Use original characters and style vocabulary.

Canvas-Native Boundary

Follow the JarvisHub Execution Model. Do not assume unlisted model choices, dialogue synthesis, or perfect one-take screen break effects. If exact dialogue or SFX cannot be generated in-video, keep them as audio/assembly notes or bind user-provided audio.

Required Intake

Collect:

• anime mode,
• original character description or reference image,
• aspect ratio: 4:3, 16:9, or 9:16,
• duration,
• output language,
• transformation/crossover direction,
• whether dialogue, BGM, SFX, OP/ED, or cover assets are needed,
• reference assets and their roles.

If the user provides a protected character reference, transform it into an original character with similar high-level traits only, removing names, logos, signature props, and distinctive protected identifiers.

Modes

Workflow

1. Lock final video spec and reference mapping.
2. Analyze character/style references.
3. Create key elements.
4. Draft storyboard.
5. Generate/bind assets.
6. Generate video clips.
7. Plan or attach audio; generate it only if this turn exposes a real audio tool.
8. Assemble and QA.

Pause after spec, storyboard, character assets, key scene/keyframe, first clip/batch, and final assembly.

Key Elements

For each character:

• daily/base form,
• transformed form,
• color identity,
• role in group,
• signature accessory/weapon,
• voice if supported.

For magical scenes:

• daily scene,
• transformation space,
• battle arena,
• villain/dark scene if any.

For crossover:

• real-world room/desk/screen,
• screen content/keyframe,
• character-in-screen form,
• character-in-reality form,
• boundary effect material.

Magical Transformation Rules

Solo transformation sequence:

1. emotional trigger,
2. object/accessory activation,
3. body becomes light silhouette,
4. ribbons/light/energy form gloves/boots/clothes,
5. hair/accessory/headpiece change,
6. hero pose hold.

Group transformation:

• shared resolve,
• synchronized accessory raise,
• quick individual transformation flashes,
• color lights converge,
• group hero pose,
• optional declaration.

Each member's color identity must stay stable. Do not give each character a full solo transformation inside a short group sequence unless the user requests a long version.

Anime Battle Rules

Emotion comes before action.

Confrontation rhythm:

• villain pressure,
• hero close-up/emotional hesitation,
• two-shot color contrast,
• emotional trigger,
• composition flips,
• final calm beat before battle.

Avoid turning every scene into random energy beams. Use eye closeups, stillness, symbolic objects, and short declarations.

Screen Crossover Rules

Choose one direction:

• character_pulls_viewer_in: screen becomes liquid/portal; character reaches out and pulls viewer in.
• viewer_pulls_character_out: hard glass breaks; viewer pulls character from screen into reality.

Material rules:

• Liquid/portal direction uses membrane, ripples, soft refraction. No glass shards.
• Glass-break direction uses hard cracks, glass fragments, pixel/electric edge. No liquid membrane.

One-take rule:

• Keep one continuous POV shot when possible.
• The screen/monitor is the physical boundary.
• The character's design must stay consistent before and after crossing.
• The viewer's hands/arms remain realistic and should not gain magical effects.

Dialogue Rules

Use exact language selected by user.

Keep lines short:

• Japanese: short conversational phrase,
• Chinese: <=10 Chinese characters where possible,
• English: <=6 words where possible.

If in-video dialogue support is unreliable, plan dialogue as an audio layer instead of embedding it into video generation.

No subtitles unless the user explicitly wants post-added subtitles.

Prompt Rules

Use visible animation/film language:

• cel shading,
• hand-painted background,
• soft glow,
• rim light,
• limited animation hold,
• dramatic close-up,
• rotating transformation camera,
• pose freeze,
• speed lines only when stylistically appropriate.

Avoid:

• protected names/logos,
• exact official costume descriptions,
• random UI text,
• over-sexualized transformation shots,
• unreadable overlong dialogue.

Screen crossover video template:

first-person POV, one continuous shot, <<<image_1>>> as confirmed start frame, [direction-specific boundary material], [character action], [viewer hand action if any], [transition into/out of screen], no subtitles, no random text, no background music unless requested.

OP/ED Segment Rules

For OP:

• introduce protagonist,
• reveal group/world,
• hint conflict,
• build to action or hero pose.

For ED:

• slower rhythm,
• reflective walk/object/sky,
• emotional afterimage,
• credits only if exact text is supplied.

Use flova-music-mv-production when uploaded music or beat timing is the main driver.

QA

Check:

• original character, not protected copy,
• color identities stable,
• transformation/battle emotional rhythm intact,
• screen boundary material does not conflict,
• POV not broken in crossover,
• no random text/subtitles,
• dialogue short and usable,
• generated clips preserve form transitions and character identity.

Fix one character sheet, keyframe, shot, or audio line at a time.