通过


你当前正在访问 Microsoft Azure Global Edition 技术文档网站。 如果需要访问由世纪互联运营的 Microsoft Azure 中国技术文档网站,请访问 https://docs.azure.cn

Content Analyzers - Analyze Binary

从输入中提取内容和字段。

POST {endpoint}/contentunderstanding/analyzers/{analyzerId}:analyzeBinary?api-version=2025-11-01
POST {endpoint}/contentunderstanding/analyzers/{analyzerId}:analyzeBinary?api-version=2025-11-01&stringEncoding={stringEncoding}&processingLocation={processingLocation}&range={range}

URI 参数

名称 必需 类型 说明
analyzerId
path True

string

minLength: 1
maxLength: 64
pattern: ^[a-zA-Z0-9._-]{1,64}$

分析仪的唯一标识符。

endpoint
path True

string (uri)

内容理解服务端点。

api-version
query True

string

minLength: 1

要用于此操作的 API 版本。

processingLocation
query

ProcessingLocation

数据可能被处理的地点。 默认是全局。

range
query

string

输入分析范围(例如)。 1-3,5,9- 文档内容使用基于1页码的页面编号,而视听内容使用整数毫秒。

stringEncoding
query

string

内容的字符串编码格式在响应中。 可能的值有“codePoint”、“utf16”和 utf8。 默认是 codePoint.“)

请求头

Media Types: "*/*"

名称 必需 类型 说明
x-ms-client-request-id

string (uuid)

请求的不透明、全局唯一的客户端生成的字符串标识符。

请求正文

Media Types: "*/*"

名称 类型 说明
input

string (binary)

文档的二进制内容需要分析。

响应

名称 类型 说明
202 Accepted

ContentAnalyzerAnalyzeOperationStatus

已接受请求进行处理,但尚未完成处理。

标头

  • Operation-Location: string
  • x-ms-client-request-id: string
Other Status Codes

Azure.Core.Foundations.ErrorResponse

意外的错误响应。

标头

x-ms-error-code: string

安全性

Ocp-Apim-Subscription-Key

基于密钥的认证,使用Azure资源的访问密钥。

类型: apiKey
在: header

EntraIdToken

Microsoft Entra ID OAuth2 使用访问令牌进行认证。

类型: oauth2
流向: accessCode
授权 URL: https://login.microsoftonline.com/common/oauth2/authorize
令牌 URL: https://login.microsoftonline.com/common/oauth2/token

作用域

名称 说明
https://cognitiveservices.azure.com/.default

示例

Analyze File

示例请求

POST {endpoint}/contentunderstanding/analyzers/myAnalyzer:analyzeBinary?api-version=2025-11-01

"RXhhbXBsZSBGaWxl"

示例响应

Operation-Location: https://myendpoint.cognitiveservices.azure.com/contentunderstanding/analyzerResults/3b31320d-8bab-4f88-b19c-2322a7f11034?api-version=2025-11-01
{
  "id": "3b31320d-8bab-4f88-b19c-2322a7f11034",
  "status": "NotStarted"
}

定义

名称 说明
AnalysisContentKind

算是媒体内容。

AnalysisResult

分析作结果。

ArrayField

从内容中提取数组字段。

AudioVisualContent

视听内容。 比如音频/wav,视频/mp4。

AudioVisualContentSegment

检测到音频/视频内容片段。

Azure.Core.Foundations.Error

错误对象。

Azure.Core.Foundations.ErrorResponse

包含错误详细信息的响应。

Azure.Core.Foundations.InnerError

包含有关错误的更具体信息的对象。 根据 Azure REST API 准则 - https://aka.ms/AzureRestApiGuidelines#handling-errors

BooleanField

从内容中提取的布尔字段。

ContentAnalyzerAnalyzeOperationStatus

提供分析作的状态详情。

ContentFieldType

字段值的语义数据类型。

ContentSpan

元素在 markdown 中的位置,以字符偏移和长度表示。

DateField

从内容中提取日期字段。

DocumentAnnotation

文档中的注释,如划线或注释。

DocumentAnnotationComment

注释与文档注释关联。

DocumentAnnotationKind

文档注释类型。

DocumentBarcode

文档中的条形码。

DocumentBarcodeKind

条形码类型。

DocumentCaption

表格或图形说明。

DocumentChartFigure

包含图表的图,如柱形图、折线图或饼图。

DocumentContent

文档内容。 例如:纯文本/纯文字,应用程序/PDF,图片/jpeg。

DocumentContentSegment

检测到的文档内容分段。

DocumentFootnote

表格或图表的脚注。

DocumentFormula

文档中的数学公式。

DocumentFormulaKind

公式类型。

DocumentHyperlink

文档中的超链接,比如网页链接或电子邮件地址。

DocumentLine

文档中的一行,由连续的单词序列组成。

DocumentMermaidFigure

包含图表的图,如流程图或网络图。

DocumentPage

文档页面的内容。

DocumentParagraph

文档中的段落,通常由一系列连续的行组成,具有共同的对齐和间距。

DocumentSection

文件中的章节。

DocumentTable

文档中的表格,由表格单元格组成,排列成矩形布局。

DocumentTableCell

文档表中的表格单元格。

DocumentTableCellKind

表单元格类型。

DocumentWord

文档中的单词,由连续的字符序列组成。 对于非空格分隔语言(如中文、日语和朝鲜语),每个字符都表示为自己的单词。

IntegerField

整数域从内容中提取。

JsonField

从内容中提取的JSON字段。

LengthUnit

长度单位由宽度、高度和源属性组成。

NumberField

从内容中提取的数字字段。

ObjectField

从内容中提取的对象字段。

OperationState

作状态

ProcessingLocation

数据可能被处理的地点。 默认是全局。

SemanticRole

段落的语义角色。

StringField

字符串字段从内容中提取。

TimeField

从内容中提取的时间场。

TranscriptPhrase

逐字稿短语。

TranscriptWord

逐字稿。

UsageDetails

使用详情。

AnalysisContentKind

算是媒体内容。

说明
document

文档内容,比如PDF、图片、文本等。

audioVisual

视听内容,如mp3、mp4等。

AnalysisResult

分析作结果。

名称 类型 默认值 说明
analyzerId

string

minLength: 1
maxLength: 64
pattern: ^[a-zA-Z0-9._-]{1,64}$

分析仪的唯一标识符。

apiVersion

string

用于分析文档的API版本。

contents AnalysisContent[]:

提取的内容。

createdAt

string (date-time)

结果创建的日期和时间。

stringEncoding

string

codePoint

内容的字符串编码格式在响应中。 可能的值有“codePoint”、“utf16”和 utf8。 默认是 codePoint.“)

warnings

Azure.Core.Foundations.Error[]

分析文件时遇到的警告。

ArrayField

从内容中提取数组字段。

名称 类型 说明
confidence

number (float)

minimum: 0
maximum: 1

预测场值的信心。

source

string

编码源用于识别字段值在内容中的位置。

spans

ContentSpan[]

与降价内容中字段值相关的跨度。

type string:

array

字段值的语义数据类型。

valueArray ContentField[]:

数组字段值。

AudioVisualContent

视听内容。 比如音频/wav,视频/mp4。

名称 类型 说明
analyzerId

string

minLength: 1
maxLength: 64
pattern: ^[a-zA-Z0-9._-]{1,64}$

生成这些内容的分析器。

cameraShotTimesMs

integer[] (int64)

视频中摄像机切换列表,以毫秒为单位的时间戳表示。 只有当returnDetails为真时才会如此。

category

string

分类内容类别。

endTimeMs

integer (int64)

内容结束时间以毫秒计。

fields

object

从内容中提取字段。

height

integer (int32)

每个视频帧的高度(如适用)以像素为单位。

keyFrameTimesMs

integer[] (int64)

视频中的关键帧列表,以毫秒为单位的时间戳表示。 只有当returnDetails为真时才会如此。

kind string:

audioVisual

内容类。

markdown

string

内容的Markdown表示。

mimeType

string

检测到内容的MIME类型。 比如应用程序/PDF、图片/jpeg等。

path

string

内容在输入中的路径。

segments

AudioVisualContentSegment[]

检测到的内容片段列表。 只有当 enableSegment 为真时。

startTimeMs

integer (int64)

内容开始时间以毫秒计。

transcriptPhrases

TranscriptPhrase[]

文字稿短语列表。 只有当returnDetails为真时才会如此。

width

integer (int32)

每个视频帧的宽度(如适用)以像素为单位。

AudioVisualContentSegment

检测到音频/视频内容片段。

名称 类型 说明
category

string

分类内容类别。

endTimeMs

integer (int64)

分段结束时间以毫秒计。

segmentId

string

段标识符。

span

ContentSpan

该段在 Markdown 内容中的跨度。

startTimeMs

integer (int64)

分段的开始时间以毫秒计。

Azure.Core.Foundations.Error

错误对象。

名称 类型 说明
code

string

服务器定义的错误代码集之一。

details

Azure.Core.Foundations.Error[]

导致此报告错误的特定错误的详细信息数组。

innererror

Azure.Core.Foundations.InnerError

包含与当前对象有关错误的更具体信息的对象。

message

string

错误的人工可读表示形式。

target

string

错误的目标。

Azure.Core.Foundations.ErrorResponse

包含错误详细信息的响应。

名称 类型 说明
error

Azure.Core.Foundations.Error

错误对象。

Azure.Core.Foundations.InnerError

包含有关错误的更具体信息的对象。 根据 Azure REST API 准则 - https://aka.ms/AzureRestApiGuidelines#handling-errors

名称 类型 说明
code

string

服务器定义的错误代码集之一。

innererror

Azure.Core.Foundations.InnerError

内部错误。

BooleanField

从内容中提取的布尔字段。

名称 类型 说明
confidence

number (float)

minimum: 0
maximum: 1

预测场值的信心。

source

string

编码源用于识别字段值在内容中的位置。

spans

ContentSpan[]

与降价内容中字段值相关的跨度。

type string:

boolean

字段值的语义数据类型。

valueBoolean

boolean

布尔场值。

ContentAnalyzerAnalyzeOperationStatus

提供分析作的状态详情。

名称 类型 说明
error

Azure.Core.Foundations.Error

描述状态为“失败”时的错误对象。

id

string

作的唯一 ID。

result

AnalysisResult

操作的结果。

status

OperationState

作状态

usage

UsageDetails

分析作的使用详情。

ContentFieldType

字段值的语义数据类型。

说明
string

纯文本。

date

日期,标准化为ISO 8601(YYYY-MM-DD)格式。

time

时间,归一化为ISO 8601(hh:mm:ss)格式。

number

数字是双精度浮点数。

integer

整数为64位带符号整数。

boolean

Boolean 值。

array

同类型的子字段列表。

object

子字段的命名列表。

json

JSON 对象。

ContentSpan

元素在 markdown 中的位置,以字符偏移和长度表示。

名称 类型 说明
length

integer (int32)

在 markdown 中,元素的长度以字符表示。

offset

integer (int32)

在 markdown 中,元素的起始位置(0 索引),用字符表示。

DateField

从内容中提取日期字段。

名称 类型 说明
confidence

number (float)

minimum: 0
maximum: 1

预测场值的信心。

source

string

编码源用于识别字段值在内容中的位置。

spans

ContentSpan[]

与降价内容中字段值相关的跨度。

type string:

date

字段值的语义数据类型。

valueDate

string (date)

日期字段值,采用ISO 8601(YYYY-MM-DD)格式。

DocumentAnnotation

文档中的注释,如划线或注释。

名称 类型 说明
author

string

注释作者。

comments

DocumentAnnotationComment[]

注释相关的评论。

createdAt

string (date-time)

注释创建的日期和时间。

id

string

注释标识符。

kind

DocumentAnnotationKind

注释类。

lastModifiedAt

string (date-time)

注释最后修改的日期和时间。

source

string

注释的位置。

spans

ContentSpan[]

与注释相关的内容跨度。

tags

string[]

与注释相关的标签。

DocumentAnnotationComment

注释与文档注释关联。

名称 类型 说明
author

string

评论作者。

createdAt

string (date-time)

创建批注的日期和时间。

lastModifiedAt

string (date-time)

评论最后修改的日期和时间。

message

string

Markdown中的评论消息。

tags

string[]

评论相关标签。

DocumentAnnotationKind

文档注释类型。

说明
highlight

高亮注释。

strikethrough

删除线注释。

underline

下划线注释。

italic

斜体注释。

bold

加粗注释。

circle

圈标注。

note

注释。

DocumentBarcode

文档中的条形码。

名称 类型 说明
confidence

number (float)

minimum: 0
maximum: 1

预测条形码的信心。

kind

DocumentBarcodeKind

条形码类型。

source

string

编码源代码,用于识别条码在内容中的位置。

span

ContentSpan

条码在降价内容中。

value

string

条形码值。

DocumentBarcodeKind

条形码类型。

说明
QRCode

二维码,定义于ISO/IEC 18004:2015。

PDF417

PDF417,依ISO 15438定义。

UPCA

GS1 12位通用产品代码。

UPCE

GS1 6位通用产品代码。

Code39

代码39条形码,定义于ISO/IEC 16388:2007。

Code128

代码128条形码,定义于ISO/IEC 15417:2007。

EAN8

GS1 8位国际商品编号(欧洲商品编号)。

EAN13

GS1 13位国际商品编号(欧洲商品编号)。

DataBar

GS1 DataBar条形码。

Code93

代码93条码,定义于ANSI/AIM BC5-1995。

Codabar

Codabar条形码,定义于ANSI/AIM BC3-1995。

DataBarExpanded

GS1 数据条扩展条码。

ITF

交错的5条码中的第2条,定义见ANSI/AIM BC2-1995。

MicroQRCode

微型二维码,定义于ISO/IEC 23941:2022。

Aztec

阿兹特克规范,定义于ISO/IEC 24778:2008。

DataMatrix

数据矩阵代码,定义于ISO/IEC 16022:2006。

MaxiCode

MaxiCode,按照ISO/IEC 16023:2000定义。

DocumentCaption

表格或图形说明。

名称 类型 说明
content

string

标题内容。

elements

string[]

字幕中的子元素。

source

string

编码源代码,标识标题在内容中的位置。

span

ContentSpan

在Markdown内容中,标题的范围。

DocumentChartFigure

包含图表的图,如柱形图、折线图或饼图。

名称 类型 默认值 说明
caption

DocumentCaption

图注。

content

图表内容用 Chart.js 配置表示。

description

string

人物描述。

elements

string[]

图的子元素,不包括任何标题或脚注。

footnotes

DocumentFootnote[]

图表脚注列表。

id

string

数字标识符。

kind string:

chart

unknown

想想那种。

role

SemanticRole

图形的语义作用。

source

string

编码源代码,用于识别图形在内容中的位置。

span

ContentSpan

降价内容中人物的范围。

DocumentContent

文档内容。 例如:纯文本/纯文字,应用程序/PDF,图片/jpeg。

名称 类型 说明
analyzerId

string

minLength: 1
maxLength: 64
pattern: ^[a-zA-Z0-9._-]{1,64}$

生成这些内容的分析器。

annotations

DocumentAnnotation[]

文档中的注释列表。 只有当enableAnnotations和returnDetails为真时。

category

string

分类内容类别。

endPageNumber

integer (int32)

内容的末页编号(1索引)。

fields

object

从内容中提取字段。

figures DocumentFigure[]:

文件中的人物列表。 只有在 enableLayout 和 returnDetails 都成立的情况下。

hyperlinks

DocumentHyperlink[]

文档中的超链接列表。 只有在returnDetails为真的情况下。

kind string:

document

内容类。

markdown

string

内容的Markdown表示。

mimeType

string

检测到内容的MIME类型。 比如应用程序/PDF、图片/jpeg等。

pages

DocumentPage[]

文件页数列表。

paragraphs

DocumentParagraph[]

文件中的段落列表。 只有当 enableOcr 和 returnDetails 都成立时。

path

string

内容在输入中的路径。

sections

DocumentSection[]

文档中的章节列表。 只有在 enableLayout 和 returnDetails 都成立的情况下。

segments

DocumentContentSegment[]

检测到的内容片段列表。 只有当 enableSegment 为真时。

startPageNumber

integer (int32)

内容的起始页码(索引1)。

tables

DocumentTable[]

文档中的表格列表。 只有在 enableLayout 和 returnDetails 都成立的情况下。

unit

LengthUnit

长度单位由宽度、高度和源属性组成。 对于图片/tiff,默认单位是像素。 对于PDF来说,默认单位是英寸。

DocumentContentSegment

检测到的文档内容分段。

名称 类型 说明
category

string

分类内容类别。

endPageNumber

integer (int32)

该段的末页编号(1索引)。

segmentId

string

段标识符。

span

ContentSpan

该段在 Markdown 内容中的跨度。

startPageNumber

integer (int32)

该段的起始页码(1索引)。

DocumentFootnote

表格或图表的脚注。

名称 类型 说明
content

string

脚注内容。

elements

string[]

脚注中的子元素。

source

string

编码源代码,标识脚注在内容中的位置。

span

ContentSpan

在折扣内容中脚注的范围。

DocumentFormula

文档中的数学公式。

名称 类型 说明
confidence

number (float)

minimum: 0
maximum: 1

对公式预测的信心。

kind

DocumentFormulaKind

公式类型。

source

string

编码源,标识公式在内容中的位置。

span

ContentSpan

公式的跨度,在 Markdown 内容中。

value

string

描述公式的 LaTex 表达式。

DocumentFormulaKind

公式类型。

说明
inline

一个嵌入在段落内容中的公式。

display

显示模式下的公式占据整行。

文档中的超链接,比如网页链接或电子邮件地址。

名称 类型 说明
content

string

超链接内容。

source

string

超链接的位置。

span

ContentSpan

在Markdown内容中超链接的范围。

url

string

超链接的网址。

DocumentLine

文档中的一行,由连续的单词序列组成。

名称 类型 说明
content

string

行文字。

source

string

编码源用于识别该行在内容中的位置。

span

ContentSpan

折扣内容中该线的宽度。

DocumentMermaidFigure

包含图表的图,如流程图或网络图。

名称 类型 默认值 说明
caption

DocumentCaption

图注。

content

string

图示内容采用 美人鱼语法表示。

description

string

人物描述。

elements

string[]

图的子元素,不包括任何标题或脚注。

footnotes

DocumentFootnote[]

图表脚注列表。

id

string

数字标识符。

kind string:

mermaid

unknown

想想那种。

role

SemanticRole

图形的语义作用。

source

string

编码源代码,用于识别图形在内容中的位置。

span

ContentSpan

降价内容中人物的范围。

DocumentPage

文档页面的内容。

名称 类型 说明
angle

number (float)

maximum: 180

内容以顺时针方向的一般方向,以度为单位(-180,180)。 只有在enableOCR成立的情况下。

barcodes

DocumentBarcode[]

页面中的条形码列表。 只有在 enableBarcode 和 returnDetails 都为真的情况下。

formulas

DocumentFormula[]

页面中的数学公式列表。 只有当 enableFormula 和 returnDetails 都为真时。

height

number (float)

页面高度。

lines

DocumentLine[]

页面中的行数列表。 只有当 enableOcr 和 returnDetails 都成立时。

pageNumber

integer (int32)

minimum: 1

页码(以1为基础)。

spans

ContentSpan[]

与 markdown 内容中页面相关的跨度。

width

number (float)

页面宽度。

words

DocumentWord[]

页面上的单词列表。 只有当 enableOcr 和 returnDetails 都成立时。

DocumentParagraph

文档中的段落,通常由一系列连续的行组成,具有共同的对齐和间距。

名称 类型 说明
content

string

段落文本。

role

SemanticRole

段落的语义角色。

source

string

编码来源,标识段落在内容中的位置。

span

ContentSpan

在降价内容中段落的长度。

DocumentSection

文件中的章节。

名称 类型 说明
elements

string[]

节的子元素。

span

ContentSpan

在降价内容中该段的跨度。

DocumentTable

文档中的表格,由表格单元格组成,排列成矩形布局。

名称 类型 说明
caption

DocumentCaption

表格说明。

cells

DocumentTableCell[]

表格中包含的单元格。

columnCount

integer (int32)

minimum: 1

表中的列数。

footnotes

DocumentFootnote[]

表格脚注列表。

role

SemanticRole

表格的语义角色。

rowCount

integer (int32)

minimum: 1

表中的行数。

source

string

编码源,标识表在内容中的位置。

span

ContentSpan

在Markdown内容中,表格的横幅。

DocumentTableCell

文档表中的表格单元格。

名称 类型 默认值 说明
columnIndex

integer (int32)

单元格的列索引。

columnSpan

integer (int32)

minimum: 1
1

此单元格跨越的列数。

content

string

表格单元的内容。

elements

string[]

表单元的子元素。

kind

DocumentTableCellKind

content

表单元格类型。

rowIndex

integer (int32)

单元格的行索引。

rowSpan

integer (int32)

minimum: 1
1

此单元格跨越的行数。

source

string

编码源代码,用于识别表格单元在内容中的位置。

span

ContentSpan

markdown 内容中表格单元的跨度。

DocumentTableCellKind

表单元格类型。

说明
content

主要内容/数据。

rowHeader

列内容描述。

columnHeader

描述 专栏内容。

stubHead

行头的描述,通常位于表格的左上角。

description

表格(部分)内容的描述。

DocumentWord

文档中的单词,由连续的字符序列组成。 对于非空格分隔语言(如中文、日语和朝鲜语),每个字符都表示为自己的单词。

名称 类型 说明
confidence

number (float)

minimum: 0
maximum: 1

预测单词的自信。

content

string

文字文字。

source

string

编码的源代码,标识该词在内容中的位置。

span

ContentSpan

在Markdown内容中,这个词的范围。

IntegerField

整数域从内容中提取。

名称 类型 说明
confidence

number (float)

minimum: 0
maximum: 1

预测场值的信心。

source

string

编码源用于识别字段值在内容中的位置。

spans

ContentSpan[]

与降价内容中字段值相关的跨度。

type string:

integer

字段值的语义数据类型。

valueInteger

integer (int64)

整数场值。

JsonField

从内容中提取的JSON字段。

名称 类型 说明
confidence

number (float)

minimum: 0
maximum: 1

预测场值的信心。

source

string

编码源用于识别字段值在内容中的位置。

spans

ContentSpan[]

与降价内容中字段值相关的跨度。

type string:

json

字段值的语义数据类型。

valueJson

JSON 字段值。

LengthUnit

长度单位由宽度、高度和源属性组成。

说明
pixel

像素单元。

inch

英寸单元。

NumberField

从内容中提取的数字字段。

名称 类型 说明
confidence

number (float)

minimum: 0
maximum: 1

预测场值的信心。

source

string

编码源用于识别字段值在内容中的位置。

spans

ContentSpan[]

与降价内容中字段值相关的跨度。

type string:

number

字段值的语义数据类型。

valueNumber

number (double)

数字字段值。

ObjectField

从内容中提取的对象字段。

名称 类型 说明
confidence

number (float)

minimum: 0
maximum: 1

预测场值的信心。

source

string

编码源用于识别字段值在内容中的位置。

spans

ContentSpan[]

与降价内容中字段值相关的跨度。

type string:

object

字段值的语义数据类型。

valueObject

object

对象字段值。

OperationState

作状态

说明
NotStarted

尚未启动操作。

Running

作正在进行中。

Succeeded

作已成功完成。

Failed

作失败。

Canceled

用户已取消作。

ProcessingLocation

数据可能被处理的地点。 默认是全局。

说明
geography

数据可能与资源在同一地理范围内处理。

dataZone

数据可以在与资源相同的数据区域内处理。

global

数据可以在全球任何 Azure 数据中心处理。

SemanticRole

段落的语义角色。

说明
pageHeader

页面顶部附近的文字。

pageFooter

页面底部附近的文字。

pageNumber

页码。

title

顶部标题描述整份文档。

sectionHeading

小标题描述文档的某一部分。

footnote

注释通常放在页面主内容之后。

formulaBlock

一组公式,通常与对齐一致。

StringField

字符串字段从内容中提取。

名称 类型 说明
confidence

number (float)

minimum: 0
maximum: 1

预测场值的信心。

source

string

编码源用于识别字段值在内容中的位置。

spans

ContentSpan[]

与降价内容中字段值相关的跨度。

type string:

string

字段值的语义数据类型。

valueString

string

字符串场值。

TimeField

从内容中提取的时间场。

名称 类型 说明
confidence

number (float)

minimum: 0
maximum: 1

预测场值的信心。

source

string

编码源用于识别字段值在内容中的位置。

spans

ContentSpan[]

与降价内容中字段值相关的跨度。

type string:

time

字段值的语义数据类型。

valueTime

string (time)

时间字段值,采用ISO 8601(hh:mm:ss)格式。

TranscriptPhrase

逐字稿短语。

名称 类型 说明
confidence

number (float)

minimum: 0
maximum: 1

预测短语的自信。

endTimeMs

integer (int64)

短语结束时间以毫秒表示。

locale

string

检测到该短语的所在地。 例如 en-US。

span

ContentSpan

折扣内容中该短语的范围。

speaker

string

演讲者索引或姓名。

startTimeMs

integer (int64)

短语的起始时间以毫秒为单位。

text

string

逐字稿文本。

words

TranscriptWord[]

短语中的单词列表。

TranscriptWord

逐字稿。

名称 类型 说明
endTimeMs

integer (int64)

单词结束时间以毫秒表示。

span

ContentSpan

在Markdown内容中,这个词的范围。

startTimeMs

integer (int64)

单词的起始时间以毫秒计。

text

string

逐字稿文本。

UsageDetails

使用详情。

名称 类型 说明
audioHours

number (float)

数小时的音频处理。

contextualizationTokens

integer (int32)

用于准备上下文、生成置信度评分、来源基础和输出格式化所消耗的上下文化代币数量。

documentPagesBasic

integer (int32)

基础层面处理的文档页数。 对于没有明确页面的文档(ex. txt、html),每3000个UTF-16字符算作一页。

documentPagesMinimal

integer (int32)

最低处理的文档页数。 对于没有明确页面的文档(ex. txt、html),每3000个UTF-16字符算作一页。

documentPagesStandard

integer (int32)

标准层面处理的文档页数。 对于没有明确页面的文档(ex. txt、html),每3000个UTF-16字符算作一页。

tokens

object

消耗的LLM和嵌入令牌数量,按型号(例如GTP 4.1)和类型(例如输入、缓存输入、输出)分组。

videoHours

number (float)

数小时的视频处理。