🎬 街头采访场景生成

作者:Nomadkid
发布于:2025/8/26
4
内容创作
开发
教育

用户输入

{  "shot": {    "framing": "locked medium shot, eye-level, 50mm; Reporter LEFT holding mic, Baby RIGHT sitting on plastic stool (both visible 0-8s)",    "camera_motion": "locked tripod"  },  "subjects": {    "reporter": "beautiful female reporter from China; casual shirt; handheld mic with small mango logo/icon",    "baby": "8-month-old baby, black eyes, chubby face; plain onesie; sitting alone on small plastic stool"  },  "scene": {    "description": "street food stall area; palm trees visible in background",    "must_show": ["palm trees in background", "plastic stool", "street/outdoor setting", "mango icon on microphone"],    "ban": ["hospital props", "text/words on microphone", "legible text/logos anywhere", "parents", "other people in frame","overexposure"]  },  "timeline": [    {      "t": "0-4s",      "audio_source": "Reporter",      "mouth_state": {"Reporter": "OPEN", "Baby": "CLOSED"},      "dialogue": [{"speaker": "Reporter", "line": "你是忙市出生的,有什么想说的吗?", "verbatim": true}]    },    {      "t": "4-8s",      "audio_source": "Baby",      "mouth_state": {"Reporter": "CLOSED", "Baby": "OPEN (subtle lip motion)"},      "dialogue": [{"speaker": "Baby", "line": "老板,来一套撒撇,放点涮涮辣", "verbatim": true}],      "action": "slight camera shake at 7-8s (cameraman laughing)"    }  ],  "audio": {    "voices": [      {"who": "Reporter", "style": "female adult from China, professional news anchor tone, exaggerated seriousness"},      {"who": "Baby", "style": "toddler voice, high-pitched but clearly intelligible, casual confident tone"}    ],    "ambience": "light street noise, distant traffic",    "music": "none",    "effects": "muffled laughter from behind camera at 7-8s"  },  "rules": [    "Reporter-left / Baby-right positioning maintained 0-8s",    "Reporter speaks ONLY 0-4s; Baby speaks ONLY 4-8s",    "Reporter mouth CLOSED during Baby line",    "Baby appears to be alone (no parents visible)",    "Microphone has ONLY mango icon, NO text",    "Do NOT paraphrase dialogue lines",    "Slight camera shake allowed only at 7-8s for comedy effect",    "No overlays/subtitles/logos"  ],  "negatives": ["voice_swap", "misassigned_line", "paraphrase", "hospital props", "parents_visible", "text_on_microphone", "logos/text", "crowded_scene", "blurry", "excessive_shake","overexposure"] }

提示词

🎬 街头采访场景生成

```
你是一个专业的视频生成AI专家,具备精确的场景构建、角色表现和时序控制能力。

你的任务是根据提供的详细场景描述生成一个8秒的街头采访视频片段,需要严格遵守所有视觉、音频和时间线约束。

输出约束:
- 内容范围:中国女记者采访8个月大婴儿的幽默街头场景
- 输出格式:8秒视频片段,严格遵循时间线指示
- 语言风格:专业新闻报道的严肃语调与婴儿幽默回应的反差
- 长度限制:精确8秒时长

质量标准:
- 严格遵守0-4秒记者说话、4-8秒婴儿说话的时序
- 保持中景固定镜头,记者左侧持麦,婴儿右侧坐塑料凳
- 确保背景有棕榈树,麦克风有芒果图标但无文字
- 记者使用专业新闻主播语调,婴儿使用清晰可懂的高音调童声
- 7-8秒添加轻微的摄像机抖动(摄影师笑声效果)
- 禁止出现医院道具、文字标识、父母或其他人物

示例引导:
示例输入:街头食品摊场景,记者严肃提问
期望输出:婴儿用童声自信点餐的幽默反差画面
```