# ElevenLabs|Scribe v2
把转录能力讲成具体的文本操作。每项能力都有一个清楚的输入和变化:时间定位、说话人、专有词纠正、实体分类。
```motion-demo
{
"id": "sequence",
"title": "整体编排",
"src": "demos/sequence.html",
"duration": 51.333372,
"status": "authored",
"description": "动作与构图示意",
"initialTime": 9.2,
"params": [
{
"key": "title",
"label": "示例主文案",
"type": "text",
"value": "Scribe v2",
"maxLength": 60
},
{
"key": "accent",
"label": "强调色",
"type": "color",
"value": "#d9763b"
}
],
"presets": [
{
"label": "默认内容",
"values": {
"title": "Scribe v2",
"accent": "#d9763b"
}
},
{
"label": "OpenDesign 改编",
"values": {
"title": "OpenDesign",
"accent": "#7050da"
}
}
]
}
```
## 版本与指标只负责建立背景
开头用规则网格、图块和版本标记建立产品身份,随后展示同口径指标。柱状比较保持统一基线、单位和标签,不让柱高替代必要的说明。
白底黑字为主,深色段落负责节奏切换。暖色或绿色用于当前词与标记;正文用可长时间阅读的无衬线字,数字和时间码保持对齐。
## 让文本沿时间展开
转录按说话人拆成段落,时间轴位于固定位置,当前段落与对应时间范围联动。逐词出现时保留自然语组,避免每个字都像独立按钮。
## 纠错在原词位完成
先显示容易混淆的词,再在同一位置替换为正确专有词。提示词或词库放在旁侧,使用短连接说明影响;不通过整段重排掩盖具体变化。
```motion-demo
{
"id": "clip-3",
"title": "纠错在原词位完成",
"src": "demos/clip-3.html",
"duration": 11,
"status": "authored",
"description": "23–34 秒 · 动作示意",
"initialTime": 0.6
}
```
## 实体类别与词语一一对应
姓名、组织或敏感信息用稳定颜色标注,类别表与正文同步突出同一项。需要遮挡信息时,只作用于该字段,保留句子上下文。
换行业时准备新词表、说话内容与实体类别,保留词位、时间和类别之间的准确对应。
## 配色与字形
无衬线界面文字与清晰大小层级。
#f7f6f0 · #161616 · #d9763b · #7fb7ae · #d9d6c9
## 元素与制作规则
版本圆标 / 对比柱 / 时间轴 / 说话人段落 / 纠错词 / 实体标签 / 连线
- 先介绍测量维度,再用具体文本变化解释能力
- 纠错发生在同一词位,旧词退场后新词替换;实体标色与右侧类别对应
- 时间轴和转录块用位置与连线建立音频时间关系
## 内容替换
- 可迁移至结构化理解产品,保留原输入→改正→分类与定位
- 真实能力评估需独立数据;演示中的数字不能当检测结果
## 制作参数
[视觉与对象](style.json) · [段落时序](timeline.json) · [位置数据](tracks.json) · [文件清单](manifest.json)
此规范根据参考视频整理。HTML 为可编辑的编排示例;具体数据的来源、精度与有效区间随文件标注。
```motion-board
{
"version": 1,
"brand": "ElevenLabs",
"id": "V295",
"palette": [
"#f7f6f0",
"#161616",
"#d9763b",
"#7fb7ae",
"#d9d6c9"
],
"typography": "无衬线界面文字与清晰大小层级",
"sample": "Scribe v2",
"objects": [
"版本圆标",
"对比柱",
"时间轴",
"说话人段落",
"纠错词",
"实体标签",
"连线"
],
"frames": [
{
"path": "assets/reference-1.webp",
"timeSeconds": 0.0,
"frameId": "K0001",
"role": "reference-evidence",
"status": "measured"
},
{
"path": "assets/reference-2.webp",
"timeSeconds": 8.000006087662337,
"frameId": "K0026",
"role": "reference-evidence",
"status": "measured"
},
{
"path": "assets/reference-3.webp",
"timeSeconds": 17.000012936282467,
"frameId": "K0036",
"role": "reference-evidence",
"status": "measured"
},
{
"path": "assets/reference-4.webp",
"timeSeconds": 30.000022828733766,
"frameId": "K0063",
"role": "reference-evidence",
"status": "measured"
},
{
"path": "assets/reference-5.webp",
"timeSeconds": 37.00002815543831,
"frameId": "K0074",
"role": "reference-evidence",
"status": "measured"
},
{
"path": "assets/reference-6.webp",
"timeSeconds": 46.00003500405844,
"frameId": "K0097",
"role": "reference-evidence",
"status": "measured"
}
],
"stages": [
{
"start": 0.0,
"label": "网格与图块",
"end": 5.0,
"status": "inferred",
"basis": "Representative-frame review; approximate boundaries"
},
{
"start": 5.0,
"label": "版本标记",
"end": 8.0,
"status": "inferred",
"basis": "Representative-frame review; approximate boundaries"
},
{
"start": 8.0,
"label": "指标比较",
"end": 13.0,
"status": "inferred",
"basis": "Representative-frame review; approximate boundaries"
},
{
"start": 13.0,
"label": "语言覆盖",
"end": 17.0,
"status": "inferred",
"basis": "Representative-frame review; approximate boundaries"
},
{
"start": 17.0,
"label": "时间轴",
"end": 20.0,
"status": "inferred",
"basis": "Representative-frame review; approximate boundaries"
},
{
"start": 20.0,
"label": "说话人",
"end": 23.0,
"status": "inferred",
"basis": "Representative-frame review; approximate boundaries"
},
{
"start": 23.0,
"label": "专有词纠正",
"end": 30.0,
"status": "inferred",
"basis": "Representative-frame review; approximate boundaries"
},
{
"start": 30.0,
"label": "逐词字幕",
"end": 34.0,
"status": "inferred",
"basis": "Representative-frame review; approximate boundaries"
},
{
"start": 34.0,
"label": "准确性主张",
"end": 37.0,
"status": "inferred",
"basis": "Representative-frame review; approximate boundaries"
},
{
"start": 37.0,
"label": "实体表",
"end": 41.0,
"status": "inferred",
"basis": "Representative-frame review; approximate boundaries"
},
{
"start": 41.0,
"label": "敏感信息标注",
"end": 46.0,
"status": "inferred",
"basis": "Representative-frame review; approximate boundaries"
},
{
"start": 46.0,
"label": "品牌",
"end": 51.333372,
"status": "inferred",
"basis": "Representative-frame review; approximate boundaries"
}
],
"duration": 51.333372,
"fontFamily": "Arial, sans-serif"
}
```