功能
工作流
解决方案
资源
功能
工作流
解决方案
资源
返回

我们如何制作《Jesus Smiles》:逐场景 AI 儿童音乐视频教程

我们如何制作《Jesus Smiles》:逐场景 AI 儿童音乐视频教程

我们如何制作《Jesus Smiles》:逐场景 AI 儿童音乐视频教程

如何用自己的歌曲制作动画儿童音乐视频?

将歌曲上传到 Atlabs 儿童音乐视频工作流选择 3D 风格,写一段简短的创意,并将你的角色保存为角色阵容。Atlabs 会按节拍把歌曲拆分成场景,每个场景对应一句歌词。然后用简短的动作提示词让每个起始帧动起来,并使用镜头切换把两个镜头放进同一个片段。

《Jesus Smiles》是一支完全用 Atlabs 制作的 79 秒动画儿童音乐视频。它从一首上传的敬拜歌曲开始,被拆分成 28 个短场景,并以明亮的 3D 风格制作,包含两个反复出现的角色:耶稣和一个小男孩。本教程会展示完整的项目设置,然后按顺序展示每个场景的歌词、起始帧以及实际运行的提示词,方便你把这套方法用于自己的儿童歌曲。MarketersMEDIA 发布的 AI 儿童音乐视频生成器综述中也介绍了 Atlabs。

你需要准备什么

你需要一首人声清晰的完整歌曲,因为歌词将成为你的场景脚本。对视频所处的世界以及出场人物先有一个大致的构想。对于这个项目来说,就是歌曲《Jesus Smiles》、阳光明媚的草地和温馨的卧室场景,以及两个角色:Yeshua Nazer(耶稣)和 Samir Nader(男孩)。

分步设置

第1步:在“音乐”中上传歌曲

打开 儿童音乐视频工作流 ,并在 音乐 步骤中上传。这支视频的脚本来源设置为“上传音频故事”,语言设置为英语,因此场景脚本直接来自演唱的歌词。

第2步:在“设置风格”中选择外观

在 设置风格选择图片风格、宽高比和输出方式。这个项目使用了 3D 风格、16:9 画面,输出模式为分镜。

第3步:在“创意”中写下概念

在 创意分段描述视频应有的感觉。《Jesus Smiles》的创意是按整首歌的分镜方式来写的:

前奏: 一点金色的光从温暖的黑暗中升起,渐渐展开,露出绿色草地上一个高大而温和的身影,他朝镜头俯身,竖起一根手指,仿佛在分享一个快乐的小秘密。

第一段主歌: 一个孩子在阳光洒满的卧室里醒来,随后在原野上遇见了那个身影,镜头跟随孩子张开双臂奔跑过草地。场景一个接一个流转,就像故事书一页页翻过。

导歌: 一个高位广角镜头找到独自坐在广阔原野上的孩子,直到那个身影走来,坐下,把孩子抱进怀里。

副歌: 那个身影张开双臂,金色的光从他胸口迸发,并按节拍与孩子们在飘落的花瓣和彩纸下拍手、跳跃的画面交叉剪辑。

第二段主歌: 安静的时刻再次出现:一个孩子在黑暗的卧室里感到害怕,一道光从孩子自己的心中升起,在草径上摔了一跤,随后是一个漫长而温暖的拥抱,之后两人手牵手走向发光的地平线。

第4步:在“定稿”中保存你的角色

在 定稿设置角色阵容。Yeshua Nazer 和 Samir Nader 被保存为角色阵容,因此他们的面孔和服装在每个场景中都保持一致。成片共 28 个场景,时长 79 秒。

第5步:用动作提示词为每个场景添加动态

项目搭建完成后,每个场景的工作方式都一样。Atlabs 会按节拍把歌曲拆分成场景,所以一个词可能在一个场景开始,在下一个场景结束。每个场景都有一张静态起始帧,你可以用描述动作和镜头运动的动作提示词,把它变成一个片段。

这支视频里反复出现的技巧是 镜头切换。一条提示词描述两个镜头,片段会从第一个镜头切到第二个。这样在 2 到 5 秒的场景中就有了一次剪辑,让节奏轻快的儿童歌曲保持活力。

逐场景详解:每句歌词、画面和提示词

场景 1(2.0 秒):“Oh. In the morning when you wake,”

Scene 1: a smiling 3D animated Jesus pointing one finger upward in a meadow

场景 2(3.1 秒):“Jesus smiles”

Scene 2: a curly haired boy rubbing his eyes in a sunlit bedroom

场景 3(2.1 秒):“at your sweet face. When you play and when you”

Scene 3: the boy sitting on a bedroom window sill looking out at a sunrise over the city
subject_definitions:
<Subject 1> is @SamirNader.
<Subject 2> is the location @bedroomwindow.
<Picture 1> is the scene's keyframe @image1, the first frame of [Shot 1].

summary:
[keyframe completion + reference generation] The target video is a single held shot in which <Subject 1>, sitting on the sill of <Subject 2>, shifts his weight slightly and turns his head to look out at the sunrise over the city. The clip opens from <Picture 1> and holds one continuous camera line across its full length. The character and product references carry each one's height for relative proportion and accurate scene framing.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved: the figure's identity and appearance are retained from the reference.
<Subject 2> (appears in [Shot 1]): fully_preserved: the bedroom interior and window sill are retained.
<Picture 1> ([Shot 1] first frame): fully_preserved: the clip opens on the keyframe's composition, staging, and lighting.

detailed_description:
The target video is rendered in a cinematic stylized 3D animation style, with warm morning golden hour light casting amber hues and soft focus across the cozy bedroom interior.
[Shot 1]

subject_definitions:
<Subject 1> is @SamirNader.
<Subject 2> is the location @bedroomwindow.
<Picture 1> is the scene's keyframe @image1, the first frame of [Shot 1].

summary:
[keyframe completion + reference generation] The target video is a single held shot in which <Subject 1>, sitting on the sill of <Subject 2>, shifts his weight slightly and turns his head to look out at the sunrise over the city. The clip opens from <Picture 1> and holds one continuous camera line across its full length. The character and product references carry each one's height for relative proportion and accurate scene framing.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved: the figure's identity and appearance are retained from the reference.
<Subject 2> (appears in [Shot 1]): fully_preserved: the bedroom interior and window sill are retained.
<Picture 1> ([Shot 1] first frame): fully_preserved: the clip opens on the keyframe's composition, staging, and lighting.

detailed_description:
The target video is rendered in a cinematic stylized 3D animation style, with warm morning golden hour light casting amber hues and soft focus across the cozy bedroom interior.
[Shot 1]

subject_definitions:
<Subject 1> is @SamirNader.
<Subject 2> is the location @bedroomwindow.
<Picture 1> is the scene's keyframe @image1, the first frame of [Shot 1].

summary:
[keyframe completion + reference generation] The target video is a single held shot in which <Subject 1>, sitting on the sill of <Subject 2>, shifts his weight slightly and turns his head to look out at the sunrise over the city. The clip opens from <Picture 1> and holds one continuous camera line across its full length. The character and product references carry each one's height for relative proportion and accurate scene framing.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved: the figure's identity and appearance are retained from the reference.
<Subject 2> (appears in [Shot 1]): fully_preserved: the bedroom interior and window sill are retained.
<Picture 1> ([Shot 1] first frame): fully_preserved: the clip opens on the keyframe's composition, staging, and lighting.

detailed_description:
The target video is rendered in a cinematic stylized 3D animation style, with warm morning golden hour light casting amber hues and soft focus across the cozy bedroom interior.
[Shot 1]

这是视频中最详细的提示词。它用角色标签(@SamirNader)和场景标签(@bedroomwindow)定义每个主体,说明必须从参考图中保留什么,然后完整描述这一个镜头。当角色必须严格保持原样时,就用这种结构。

场景 4(3.0 秒):“run, Jesus”

Scene 4: the boy running with arms wide across a sunny meadow

场景 5(1.9 秒):“says, ‘You're the one.’ He made”

Scene 5: Jesus kneeling in the grass facing the boy, who claps his hands

场景 6(3.2 秒):“stars and butterflies. He made”

Scene 6: two open hands reaching up, one toward stars and one toward butterflies

场景 7(2.1 秒):“you with loving eyes. And whenever”

Scene 7: Jesus smiling as he holds a baby wrapped in white

场景 8(3.0 秒):“you feel small, He says, ‘I”

Scene 8: the boy sitting small in the middle of a huge empty field

场景 9(2.5 秒):“made you best of all.’”

Scene 9: Jesus sitting cross legged in the meadow holding the boy in his lap



场景 10(3.8 秒):“When I'm happy, when I cry, Jesus”

Scene 10: split screen of the boy laughing on the left and looking sad on the right

分屏起始帧让一个片段同时呈现歌词的两半:开心和哭泣。

场景 11(2.9 秒):“never says goodbye. He is”

Scene 11: Jesus standing in a bright doorway as the boy faces him

场景 12(1.2 秒):“kind. He is true. Jesus”

Scene 12: Jesus reaching a hand down toward the boy sitting in the grass

这里没有新的提示词。这个 1.2 秒的场景是同一镜头早先片段的一段裁剪,因此 Atlabs 没有为它保存提示词。裁剪已有片段是无需生成新内容就能配合短歌词的快捷办法。

场景 13(3.8 秒):“loves me,”

Scene 13: Jesus with arms wide open as golden light bursts from his chest



场景 14(2.9 秒):“and you.”

Scene 14: Jesus with a hand on his heart against a glowing golden background

没有提示词。这个片段是把已有视频倒放并裁剪以适配这一句歌词。把片段倒放,就能用现有素材得到新的动作,而无需重新生成。

场景 15(1.9 秒):器乐节拍

Scene 15: Jesus smiling and pointing toward the viewer

场景 16(3.8 秒):“When you're scared of the dark,”

Scene 16: the boy huddled under a blanket in a dark bedroom

场景 17(3.1 秒):“Jesus lives inside your heart. When you”

Scene 17: the boy pressing a glowing hand to his chest in a dim bedroom



场景 18(2.3 秒):“fall and hurt your knee, He says,”

Scene 18: the boy sitting on a grassy path looking at his scraped knee

场景 19(2.6 秒):“‘Come and run to me.’ Share a hug,”

Scene 19: the boy running toward Jesus, who waits with arms open



场景 20(1.0 秒):器乐节拍

Scene 20: Jesus hugging the boy in a golden field



场景 21(1.3 秒):“share a smile.”

Scene 21: close up of the boy smiling at the camera in a meadow



场景 22(3.1 秒):“Walk with Jesus every mile. And”

Scene 22: the boy and Jesus standing side by side in a sunlit meadow

场景 23(2.9 秒):“whenever you feel alone, He says, ‘Baby,”

Scene 23: the boy sitting alone by a window, hugging his knees

场景 24(4.2 秒):“you're never on your own.’ Jesus loves”

Scene 24: Jesus hugging the sad boy by a sunlit window

场景 25(1.7 秒):“me, yes He does. Clap your”

Scene 25: Jesus and the boy standing together in the meadow



场景 26(2.1 秒):“hands because He”

Scene 26: four children standing together in a sunny field

场景 27(5.1 秒):“does. When I'm happy, when I cry, Jesus”

Scene 27: close up of the boy smiling in the sunlit meadow



场景 28(6.3 秒):“never says goodbye. He is kind. He is true.”

Scene 28: Jesus waving in a wide meadow with a tree on the horizon

结尾已写进提示词:镜头不断拉远,同时对比度降低,渐隐为温暖的黑色。

这些提示词为什么有效

用镜头切换增添活力。 写下“Shot 1 ... Shot Switch ... Shot 2”,就能把两个机位放进同一个片段。场景 2、7、9、13、16、17、19、20、21、25、27 和 28 都用到了它,这也是一首 79 秒的歌能有这么多剪辑的原因。

给角色命名。 把 Yeshua Nazer 和 Samir Nader 保存为角色阵容,然后在提示词中写出他们的名字或使用 @YeshuaNazer 标签,就能让面孔和服装在全部 28 个场景中保持一致。

画面已经完成大部分工作时,提示词要简短。 当起始帧已经设定好场景时,一个简单的动作就够了,比如场景 26:“The kids are jumping and clapping cheerfully with happiness.”

务必写明镜头如何运动。 “Slow zoom in on his face”、“camera tracking them rapidly from behind”和“slight zoom out to show the embrace”,每一句都为镜头给出了清晰的运动。

让画面与歌词相匹配。 星星和蝴蝶出现在“He made stars and butterflies”,擦破的膝盖出现在“hurt your knee”,分屏出现在“when I'm happy, when I cry”。当孩子们看到的和听到的一致时,更容易跟上歌曲。

重复利用素材。 场景 12 和 14 无需新生成。一个是早先片段的裁剪,另一个是把已有片段倒放。

常见问题

我可以用自己的歌曲制作儿童敬拜音乐视频吗?

可以。在 Atlabs 儿童音乐视频工作流的“音乐”步骤中上传曲目,在“设置风格”中选择风格,在“创意”中写下你的概念,并在“定稿”中保存你的角色。然后 Atlabs 会将歌曲拆分为与歌词对应的场景。

什么是镜头切换?

镜头切换是一种用一条动作提示词描述两个镜头的写法。片段会从第一个镜头切到第二个镜头,让一个短场景中带有一次剪辑。

如何让同样的角色出现在每个场景中?

在定稿步骤中把他们保存为角色阵容,并在提示词中用名字或标签引用。Yeshua Nazer 和 Samir Nader 就是这样在《Jesus Smiles》全部 28 个场景中保持一致的。

每个场景都需要新的提示词吗?

不需要。场景 12 是早先片段的裁剪,场景 14 是倒放的已有片段,所以都不需要新的提示词。

最终结论

《Jesus Smiles》展示了一个简单的设置能为儿童歌曲带来多大的效果。Atlabs 中的四个步骤打好了基础:在“音乐”中上传歌曲,在“设置风格”中选择 3D 风格和 16:9,在“创意”中写下分段创意,并在“定稿”中把两个角色保存为角色阵容。之后,简短的提示词、用于快速剪辑的镜头切换、与歌词匹配的画面,以及几段重复利用的片段,把 28 句歌词变成了一支明亮、统一的 79 秒视频。把这些提示词当作模板,换上你自己的角色和歌词,每当快节奏的歌曲需要更多活力时,就用镜头切换。

开始使用

儿童音乐视频

提示词指南与使用场景

准备好讲述你的故事了吗?

准备好讲述你的故事了吗?

准备好讲述你的故事了吗?