纸拼贴风桃太郎动画
- 播放
- 7,371
- 点赞
- 119
- 评论
- 5
- 收藏
- 147
一段45秒的纸拼贴风格桃太郎动画由Claude Opus 5.5根据详细结构化提示词制作，图像使用Codex生成，旁白与音效使用ElevenLabs完成。
完整提示词2,334 字符
Make a 45-second paper-collage animation of Momotaro (the Japanese folktale). [Look] A pop-up picture book: the book opens and paper pieces stand up one by one. Washi paper cut-out (kirie) texture. The 3D effect comes from images and code only. Do not use any video-generation AI. [How to build it] 1. First, write the script and show it to me (8 scenes, 4-9 seconds each; the only on-screen text is short onomatopoeia such as "donburako"). 2. Once I approve the script, write the image-generation prompts: - one texture reference image - characters and backgrounds as sheets of separate paper cut-outs on transparent backgrounds 3. I will generate the images and hand them to you. Then animate them in Python: - make it look 3D with five tricks: depth, parallax, blur, shadows, and motion blur - 1920x1080, 24 fps - for each scene, show me 4 still frames first so I can check 4. Add narration and background music. [Rules] - Do not wobble or shake the pictures. Movement is only slow pushes in and pulls back. - Don't give me English-only prompts; add an explanation in my language too. @ckrtttt @nakazakifam ありがとうございます。 以下のようなプロンプトをClaudeに入力すると作れるはずです。 役割分担としては、 ・Codexは画像生成 ・ElevenLabsはBGM・ナレーション生成 です。 ClaudeとMCP連携し、Claude上で依頼する感じです。 Make a 45-second paper-collage animation of Momotaro (the Japanese folktale). [Look] A pop-up picture book: the book opens and paper pieces stand up one by one. Washi paper cut-out (kirie) texture. The 3D effect comes from images and code only. Do not use any video-generation AI. [How to build it] 1. First, write the script and show it to me (8 scenes, 4-9 seconds each; the only on-screen text is short onomatopoeia such as "donburako"). 2. Once I approve the script, write the image-generation prompts: - one texture reference image - characters and backgrounds as sheets of separate paper cut-outs on transparent backgrounds 3. I will generate the images and hand them to you. Then animate them in Python: - make it look 3D with five tricks: depth, parallax, blur, shadows, and motion blur - 1920x1080, 24 fps - for each scene, show me 4 still frames first so I can check 4. Add narration and background music. [Rules] - Do not wobble or shake the pictures. Movement is only slow pushes in and pulls back. - Don't give me English-only prompts; add an explanation in my language too.
出处：作者在原帖下的回复。提示词保留原文语言，未做改写。
需自备参考素材：原作除了文字提示词，还用了图片、视频、音频、文档或代码库作为输入，本站未收录这些素材。只复制提示词得不到同样的结果。
作者提到的工具
CodexElevenLabs
原帖：https://x.com/tanabe_fragm/status/2110931024469540979
全部作品数据开源在 GitHub · awesome-opus-5.5-video。
模型归属以作者自述为准，本站未独立复现。播放量为 2026-10-17 的快照。作品版权归原作者，修改署名或下架请在 X 私信 @GoSailGlobal。
