Schnellstart: Verwenden von Azure Content Understanding in Foundry Tools REST API

In dieser Schnellstartanleitung erfahren Sie, wie Sie die REST-API für Inhaltsverständnis verwenden, um strukturierte Daten aus multimodalen Inhalten in Dokument-, Bild-, Audio- und Videodateien abzurufen.

Voraussetzungen

Um zu beginnen, benötigen Sie ein aktives Azure-Abonnement. Falls Sie kein Azure-Konto haben, können Sie kostenlos eines erstellen.
Wenn Sie über Ihr Azure-Abonnement verfügen, erstellen Sie eine im Azure-Portal eine Microsoft Foundry-Ressource. Achten Sie darauf, sie in einer unterstützten Region zu erstellen.
Richten Sie Standardmodellbereitstellungen für Ihre Ressource für Inhaltsverständnis ein. Das Festlegen von Standardwerten erstellt eine Verbindung mit den Foundry-Modellen, die Sie für Inhaltsverständnisanforderungen verwenden. Wählen Sie eine der folgenden Methoden:
- Portal
- REST-API
1. Zur Seite " Einstellungen für inhaltsverständnis" wechseln
2. Wählen Sie oben links die Schaltfläche "+ Ressource hinzufügen" aus.
3. Wählen Sie die Foundry-Ressource aus, die Sie verwenden möchten, und klicken Sie auf "Weiter", und klicken Sie dann auf "Speichern".
  - Stellen Sie sicher, dass Sie "Automatische Bereitstellung für erforderliche Modelle aktivieren, wenn keine Standardwerte verfügbar sind" aktiviert ist. Dadurch wird sichergestellt, dass Ihre Ressource vollständig mit den erforderlichen Modellen GPT-4.1, GPT-4.1-Mini und Text-Embedding-3-Large eingerichtet ist. Für unterschiedliche vorkonfigurierte Analysegeräte sind unterschiedliche Modelle erforderlich.
Indem Sie diese Schritte ausführen, richten Sie eine Verbindung zwischen Inhaltsverständnis- und Foundry-Modellen in Ihrer Foundry-Ressource ein.
1. Erstellen Sie Foundry Models-Bereitstellungen der Typen GPT-4.1, GPT-4.1-mini und text-embedding-3-large in Ihrer Foundry-Ressource. Ausführliche Informationen zum Bereitstellen dieser Modelle finden Sie unter Erstellen von Modellbereitstellungen im Microsoft Foundry-Portal. Unterschiedliche vorgefertigte Analysegeräte erfordern unterschiedliche Modelle, daher müssen Sie alle drei Bereitstellen.
2. Definieren Sie Standardmodellbereitstellungen auf Ressourcenebene.
  
  Bevor Sie den folgenden cURL-Befehl ausführen, nehmen Sie die folgenden Änderungen an der HTTP-Anforderung vor:
  - Ersetzen Sie {endpoint} und {key} durch die entsprechenden Werte aus Ihrer Foundry-Instanz im Azure-Portal.
  - Ersetzen Sie {myGPT41Deployment}, {myGPT41MiniDeployment} und {myEmbeddingDeployment} durch die tatsächlichen Bereitstellungsnamen Ihres Modells aus Ihrer Foundry-Ressource.
```
curl -i -X PATCH "{endpoint}/contentunderstanding/defaults?api-version=2025-11-01" \
  -H "Ocp-Apim-Subscription-Key: {key}" \
  -H "Content-Type: application/json" \
  -d '{
        "modelDeployments": {
          "gpt-4.1": "{myGPT41Deployment}",
          "gpt-4.1-mini": "{myGPT41MiniDeployment}",
          "text-embedding-3-large": "{myEmbeddingDeployment}"
        }
      }'
```
Indem Sie diese Schritte ausführen, richten Sie eine Verbindung zwischen Inhaltsverständnis- und Foundry-Modellen in Ihrer Foundry-Ressource ein.
In diesem Handbuch verwenden wir das Befehlszeilentool cURL. Wenn sie nicht installiert ist, können Sie die entsprechende Version für Ihre Entwicklungsumgebung herunterladen .

Erste Schritte mit einem vorgefertigten Analyzer

Analysatoren definieren, wie Ihre Inhalte verarbeitet werden, und die Erkenntnisse, die extrahiert werden. Wir bieten vorgefertigte Analysegeräte für gängige Anwendungsfälle an. Sie können vorgefertigte Analysegeräte an Ihre spezifischen Anforderungen und Anwendungsfälle anpassen. Diese Schnellstartanleitung verwendet vorgefertigte Rechnungs-, Bild-, Audio- und Videoanalysatoren, um Ihnen bei den ersten Schritten zu helfen.

Senden einer Datei für die Analyse

Bevor Sie den folgenden cURL-Befehl ausführen, nehmen Sie die folgenden Änderungen an der HTTP-Anforderung vor:

Ersetzen Sie {endpoint} und {key} durch die entsprechenden Werte aus Ihrer Foundry-Instanz im Azure-Portal.

POST-Anforderung

In diesem Beispiel wird die prebuilt-invoice Analyse verwendet, um strukturierte Daten aus einem Rechnungsdokument zu extrahieren.

curl -i -X POST "{endpoint}/contentunderstanding/analyzers/prebuilt-invoice:analyze?api-version=2025-11-01" \
  -H "Ocp-Apim-Subscription-Key: {key}" \
  -H "Content-Type: application/json" \
  -d '{
        "inputs":[{"url": "https://github.com/Azure-Samples/azure-ai-content-understanding-python/raw/refs/heads/main/data/invoice.pdf"}]
      }'

In diesem Beispiel wird die prebuilt-imageSearch Analyse verwendet, um eine Beschreibung des Bilds zu generieren.

curl -i -X POST "{endpoint}/contentunderstanding/analyzers/prebuilt-imageSearch:analyze?api-version=2025-11-01" \
  -H "Ocp-Apim-Subscription-Key: {key}" \
  -H "Content-Type: application/json" \
  -d '{
        "inputs":[
          {
            "url": "https://github.com/Azure-Samples/azure-ai-content-understanding-python/raw/refs/heads/main/data/pieChart.jpg"
          }          
        ]
      }'

In diesem Beispiel wird die prebuilt-audioSearch Analyse verwendet, um das Audiotranskript zu extrahieren, eine Zusammenfassung zu generieren und Lautsprecherbeschriftungen durchzuführen.

curl -i -X POST "{endpoint}/contentunderstanding/analyzers/prebuilt-audioSearch:analyze?api-version=2025-11-01" \
  -H "Ocp-Apim-Subscription-Key: {key}" \
  -H "Content-Type: application/json" \
  -d '{
        "inputs":[
          {
            "url": "https://github.com/Azure-Samples/azure-ai-content-understanding-python/raw/refs/heads/main/data/audio.wav"
          }          
        ]
      }'

In diesem Beispiel wird die prebuilt-videoSearch Analyse verwendet, um Keyframes, Transkriptionen und Kapitelsegmente aus Video zu extrahieren.

curl -i -X POST "{endpoint}/contentunderstanding/analyzers/prebuilt-videoSearch:analyze?api-version=2025-11-01" \
  -H "Ocp-Apim-Subscription-Key: {key}" \
  -H "Content-Type: application/json" \
  -d '{
        "inputs":[
          {
            "url": "https://github.com/Azure-Samples/azure-ai-content-understanding-python/raw/refs/heads/main/data/FlightSimulator.mp4"
          }          
        ]
      }'

POST-Antwort

Der Antwortheader enthält ein Operation-Location Feld, das Sie zum Abrufen der Ergebnisse des asynchronen Analysevorgangs verwenden.

HTTP/1.1 202 Accepted
Transfer-Encoding: chunked
Content-Type: application/json
request-id: aaa-bbb-ccc-ddd
x-ms-request-id: aaa-bbb-ccc-ddd
Operation-Location: {endpoint}/contentunderstanding/analyzerResults/{request-id}?api-version=2025-11-01
api-supported-versions: 2024-12-01-preview,2025-05-01-preview,2025-11-01
x-envoy-upstream-service-time: 800
apim-request-id: {request-id}
Strict-Transport-Security: max-age=31536000; includeSubDomains; preload
x-content-type-options: nosniff
x-ms-region: West US
Date: Fri, 31 Oct 2025 05:30:17 GMT
Connection: close

Von Bedeutung

Kopieren Sie die Operation-Location-URL aus dem Antwortheader. Sie verwenden diese URL im nächsten Schritt, um die Analyseergebnisse abzurufen.

Abrufen des Analyseergebnisses

Verwenden Sie die Operation-Location von der POST Antwort, und rufen Sie das Ergebnis der Analyse ab.

GET-Anforderung

curl -i -X GET "{endpoint}/contentunderstanding/analyzerResults/{request-id}?api-version=2025-11-01" \
  -H "Ocp-Apim-Subscription-Key: {key}"

GET-Antwort

Die JSON-Antwort 200 (OK) enthält ein Feld status, das den Status des Vorgangs angibt. Wenn der Vorgang nicht abgeschlossen ist, lautet der Wert von status entweder Running oder NotStarted. In solchen Fällen sollten Sie die GET Anforderung erneut senden, entweder manuell oder über ein Skript. Warten Sie zwischen den Aufrufen mindestens eine Sekunde.

{
  "id": "ce05fb5a-579e-4f0b-afb5-3532bcddeaee",
  "status": "Succeeded",
  "result": {
    "analyzerId": "prebuilt-invoice",
    "apiVersion": "2025-11-01",
    "createdAt": "2025-11-13T20:04:55Z",
    "warnings": [],
    "contents": [
      {
        "path": "input1",
        "markdown": "CONTOSO LTD.\n\n\n# INVOICE\n\nContoso Headquarters\n123 456th St\nNew York, NY, 10001\n\nINVOICE: INV-100\n\nINVOICE DATE: 11/15/2019\n\nDUE DATE: 12/15/2019\n\nCUSTOMER NAME: MICROSOFT CORPORATION...",
        "fields": {
          "AmountDue": {
            "type": "object",
            "valueObject": {
              "Amount": {
                "type": "number",
                "valueNumber": 610,
                "spans": [
                  {
                    "offset": 1522,
                    "length": 7
                  }
                ],
                "confidence": 0.773,
                "source": "D(1,7.3628,8.0459,7.9272,8.0459,7.9272,8.2070,7.3628,8.2070)"
              },
              "CurrencyCode": {
                "type": "string",
                "valueString": "USD"
              }
            }
          },
          "BalanceForward": {
            "type": "object",
            "valueObject": {
              "Amount": {
                "type": "number",
                "valueNumber": 500,
                "spans": [
                  {
                    "offset": 1474,
                    "length": 7
                  }
                ],
                "confidence": 0.901,
                "source": "D(1,7.3628,7.7445,7.9278,7.7467,7.9272,7.9092,7.3622,7.9070)"
              },
              "CurrencyCode": {
                "type": "string",
                "valueString": "USD"
              }
            }
          },
          "BillingAddress": {
            "type": "string",
            "valueString": "123 Bill St, Redmond WA, 98052",
            "spans": [
              {
                "offset": 325,
                "length": 12
              },
              "..."
            ],
            "confidence": 0.712,
            "source": "D(1,0.5805,3.9471,1.2858,3.9478,1.2856,4.1115,0.5803,4.1108);..."
          },
          "BillingAddressRecipient": {
            "type": "string",
            "valueString": "Microsoft Finance",
            "spans": [
              {
                "offset": 307,
                "length": 17
              }
            ],
            "confidence": 0.815,
            "source": "D(1,0.5734,3.7392,1.8060,3.7521,1.8043,3.9201,0.5717,3.9072)"
          },
          "CountryRegion": {
            "type": "string",
            "valueString": "USA"
          },
          "CustomerAddress": {
            "type": "string",
            "valueString": "123 Other St, Redmond WA, 98052",
            "spans": [
              "..."
            ],
            "confidence": 0.744,
            "source": "..."
          },
          "CustomerAddressRecipient": {
            "type": "string",
            "valueString": "Microsoft Corp",
            "spans": [
              "..."
            ],
            "confidence": 0.437,
            "source": "..."
          },
          "CustomerId": {
            "type": "string",
            "valueString": "CID-12345",
            "spans": [
              "..."
            ],
            "confidence": 0.936,
            "source": "..."
          },
          "CustomerName": {
            "type": "string",
            "valueString": "MICROSOFT CORPORATION",
            "spans": [
              "..."
            ],
            "confidence": 0.46,
            "source": "..."
          },
          "CustomerTaxId": {
            "type": "string",
            "confidence": 0.912
          },
          "DueDate": {
            "type": "date",
            "valueDate": "2019-12-15",
            "spans": [
              "..."
            ],
            "confidence": 0.97,
            "source": "..."
          },
          "InvoiceDate": {
            "type": "date",
            "valueDate": "2019-11-15",
            "spans": [
              "..."
            ],
            "confidence": 0.939,
            "source": "..."
          },
          "InvoiceId": {
            "type": "string",
            "valueString": "INV-100",
            "spans": [
              "..."
            ],
            "confidence": 0.733,
            "source": "..."
          },
          "LineItems": {
            "type": "array",
            "valueArray": [
              {
                "type": "object",
                "valueObject": {
                  "Date": {
                    "type": "date",
                    "valueDate": "2021-03-04",
                    "spans": [
                      "..."
                    ],
                    "confidence": 0.894,
                    "source": "..."
                  },
                  "Description": {
                    "type": "string",
                    "valueString": "Consulting Services",
                    "spans": [
                      "..."
                    ],
                    "confidence": 0.589,
                    "source": "..."
                  },
                  "ProductCode": {
                    "type": "string",
                    "valueString": "A123",
                    "spans": [
                      "..."
                    ],
                    "confidence": 0.879,
                    "source": "..."
                  },
                  "Quantity": {
                    "type": "number",
                    "valueNumber": 2,
                    "spans": [
                      "..."
                    ],
                    "confidence": 0.939,
                    "source": "..."
                  },
                  "QuantityUnit": {
                    "type": "string",
                    "valueString": "hours",
                    "spans": [
                      "..."
                    ],
                    "confidence": 0.85,
                    "source": "..."
                  },
                  "TaxAmount": {
                    "type": "object",
                    "valueObject": {
                      "Amount": {
                        "type": "number",
                        "valueNumber": 6,
                        "spans": [
                          "..."
                        ],
                        "confidence": 0.522,
                        "source": "..."
                      },
                      "CurrencyCode": {
                        "type": "string",
                        "valueString": "USD"
                      }
                    }
                  },
                  "TaxRate": {
                    "type": "number",
                    "confidence": 0.915
                  },
                  "TotalAmount": {
                    "type": "object",
                    "valueObject": {
                      "Amount": {
                        "type": "number",
                        "valueNumber": 60,
                        "spans": [
                          "..."
                        ],
                        "confidence": 0.972,
                        "source": "..."
                      },
                      "CurrencyCode": {
                        "type": "string",
                        "valueString": "USD"
                      }
                    }
                  },
                  "UnitPrice": {
                    "type": "object",
                    "valueObject": {
                      "Amount": {
                        "type": "number",
                        "valueNumber": 30,
                        "spans": [
                          "..."
                        ],
                        "confidence": 0.97,
                        "source": "..."
                      },
                      "CurrencyCode": {
                        "type": "string",
                        "valueString": "USD"
                      }
                    }
                  }
                }
              },
              "... (2 additional line items)"
            ]
          }
          /*additional fields omitted*/
        },
        "kind": "document",
        "startPageNumber": 1,
        "endPageNumber": 1,
        "unit": "inch",
        "pages": [
          {
            "pageNumber": 1,
            "angle": 0,
            "width": 8.5,
            "height": 11,
            "words": [
              "... (words omitted for brevity)"
            ],
            "selectionMarks": [],
            "lines": [
              "... (lines omitted for brevity)"
            ],
            "barcodes": [],
            "formulas": []
          }
        ],
        "tables": [
          "... (tables omitted for brevity)"
        ],
        "keyValuePairs": [
          "... (key-value pairs omitted for brevity)"
        ],
        "analyzerId": "prebuilt-invoice",
        "mimeType": "application/pdf"
      }
    ]
  },
  "usage": {
    "documentStandardPages": 1,
    "contextualizationTokens": 2345,
    "tokens": {
      "gpt-4.1-mini-input": 1234,
      "gpt-4.1-mini-output": 567
    }
  }
}

{
  "id": "fe6bb69d-1d6b-4698-a50c-ce798bacdd95",
  "status": "Succeeded",
  "result": {
    "analyzerId": "prebuilt-imageSearch",
    "apiVersion": "2025-11-01",
    "createdAt": "2025-11-13T20:11:50Z",
    "warnings": [],
    "contents": [
      {
        "path": "input1",
        "markdown": "![image](pages/1)\n",
        "fields": {
          "Summary": {
            "type": "string",
            "valueString": "The image is a 3D pie chart illustrating the distribution of hours spent in four different ranges. The largest segment, colored purple, represents 60+ hours at 37.8%. The second largest segment, in teal, represents 50-60 hours at 36.6%. The red segment shows 40-50 hours at 18.9%, and the smallest orange segment represents 1-39 hours at 6.7%. Each segment is labeled with its corresponding range and percentage."
          }
        },
        "kind": "document",
        "startPageNumber": 1,
        "endPageNumber": 1,
        "unit": "pixel",
        "pages": [
          {
            "pageNumber": 1,
            "spans": []
          }
        ],
        "analyzerId": "prebuilt-imageSearch",
        "mimeType": "image/jpeg"
      }
    ]
  },
  "usage": {
    "contextualizationTokens": 1000,
    "tokens": {
      "gpt-4.1-mini-input": 199,
      "gpt-4.1-mini-output": 106
    }
  }
}

{
  "id": "<request-id>",
  "status": "Succeeded",
  "result": {
    "analyzerId": "prebuilt-audioSearch",
    "apiVersion": "2025-11-01",
    "createdAt": "YYYY-MM-DDTHH:MM:SSZ",
    "stringEncoding": "utf8",
    "warnings": [],
    "contents": [
      {
        "path": "input1",
        "markdown": "# Audio: 00:00.000 => 01:54.670\n\nTranscript\n```\nWEBVTT\n\n00:00.080 --> 00:02.160\n<v Speaker 1>Thank you for calling Woodgrove Travel...",
        "fields": {
          "Summary": {
            "type": "string",
            "valueString": "John Smith contacted Woodgrove Travel to report a negative experience with his flight from New York City to Los Angeles..."
          }
        },
        "kind": "audioVisual",
        "startTimeMs": 0,
        "endTimeMs": 114670,
        "analyzerId": "prebuilt-audioSearch",
        "mimeType": "audio/wav"
      }
    ]
  },
  "usage": {
		"audioHours": 0.032,
    "contextualization": 3194.445
    "tokens": {
      "gpt-4.1-input": 1234, 
      "gpt-4.1-output": 2345,
      "text-embedding-3-large": 3456 
    }
 
	}
}

{
  "id": "2689a699-fa3a-4ddf-9a27-c34ceaa6c597",
  "status": "Succeeded",
  "result": {
    "analyzerId": "prebuilt-videoSearch",
    "apiVersion": "2025-11-01",
    "createdAt": "2025-11-13T16:11:17Z",
    "warnings": [],
    "contents": [
      {
        "markdown": "# Video: 00:00.733 => 00:15.467\nWidth: 1080\nHeight: 608\n\nTranscript\n```\nWEBVTT\n\n00:01.360 --> 00:06.640\n<Speaker 1>When it comes to the neural TTS, in order to get a good voice, it's better to have good data.\n\n00:07.120 --> 00:13.320\n<Speaker 2>To achieve that, we build a universal TTS model based on 3,000 hours of data.\n\n00:13.440 --> 00:23.680\n<Speaker 1>We actually accumulated tons of the data so that this universal model is able to capture the nuance of the audio and generate a more natural voice for the algorithm.\n```\n\nKey Frames\n- 00:00.733 ![](keyFrame.733.jpg)\n- 00:02.067 ![](keyFrame.2067.jpg)\n...\n- 00:14.833 ![](keyFrame.14833.jpg)\n- 00:15.467 ![](keyFrame.15467.jpg)",
        "fields": {
          "Summary": {
            "type": "string",
            "valueString": "The video opens with a Flight Simulator logo alongside Microsoft Azure AI branding, followed by an interview with a man discussing the importance of good data for neural text-to-speech (TTS) technology. He explains the creation of a universal TTS model trained on 3,000 hours of data to capture audio nuances and produce natural-sounding voices. Visuals include audio waveform displays and shots of data centers and server rooms, emphasizing the technological infrastructure behind the TTS model."
          }
        },
        "kind": "audioVisual",
        "startTimeMs": 733,
        "endTimeMs": 15467,
        "width": 1080,
        "height": 608,
        "KeyFrameTimesMs": [
          733,
          2067
          /*... (14 additional keyframes)*/
        ],
        "transcriptPhrases": [
          {
            "speaker": "Speaker 1",
            "startTimeMs": 1360,
            "endTimeMs": 6640,
            "text": "When it comes to the neural TTS, in order to get a good voice, it's better to have good data.",
            "confidence": 0.937,
            "words": [
              {
                "startTimeMs": 1360,
                "endTimeMs": 1600,
                "text": "When"
              },
              {
                "startTimeMs": 1600,
                "endTimeMs": 1760,
                "text": "it"
              }
              /*... (18 additional words)*/
            ],
            "locale": "en-US"
          },
          {
            "speaker": "Speaker 2",
            "startTimeMs": 7120,
            "endTimeMs": 13320,
            "text": "To achieve that, we build a universal TTS model based on 3,000 hours of data.",
            "confidence": 0.937,
            "words": [
              {
                "startTimeMs": 7120,
                "endTimeMs": 7360,
                "text": "To"
              },
              {
                "startTimeMs": 7560,
                "endTimeMs": 7880,
                "text": "achieve"
              }
              /*... (13 additional words)*/
            ],
            "locale": "en-US"
          },
          {
            "speaker": "Speaker 1",
            "startTimeMs": 13440,
            "endTimeMs": 23680,
            "text": "We actually accumulated tons of the data so that this universal model is able to capture the nuance of the audio and generate a more natural voice for the algorithm.",
            "confidence": 0.937,
            "words": [
              {
                "startTimeMs": 13440,
                "endTimeMs": 13600,
                "text": "We"
              },
              {
                "startTimeMs": 13600,
                "endTimeMs": 14000,
                "text": "actually"
              }
              /*... (28 additional words)*/
            ],
            "locale": "en-US"
          }
        ],
        "cameraShotTimesMs": [
          1467,
          3233
          /*... (13 additional camera shots)*/
        ],
        "mimeType": "video/x-m4v"
      },
      {
        "markdown": "# Video: 00:15.467 => 00:23.100\nWidth: 1080\nHeight: 608\n\n\n\nKey Frames\n- 00:15.467 ![](keyFrame.15467.jpg)\n- 00:16.933 ![](keyFrame.16933.jpg)\n...\n- 00:22.367 ![](keyFrame.22367.jpg)\n- 00:23.100 ![](keyFrame.23100.jpg)",
        "fields": {
          "Summary": {
            "type": "string",
            "valueString": "The video transitions to scenic aerial views from the Flight Simulator, showcasing detailed landscapes including coastlines, mountains, and castles. This segment highlights the realistic graphics and immersive experience of the Flight Simulator, demonstrating the integration of advanced AI technologies to enhance the simulation."
          }
        },
        "kind": "audioVisual",
        "startTimeMs": 15467,
        "endTimeMs": 23100,
        "width": 1080,
        "height": 608,
        "KeyFrameTimesMs": [
          15467,
          16933
          /*... (7 additional keyframes)*/
        ],
        "transcriptPhrases": [],
        "cameraShotTimesMs": [
          1467,
          3233
          /*... (13 additional camera shots)*/
        ]
      },
      {
        "markdown": "# Video: 00:23.100 => 00:43.233\nWidth: 1080\nHeight: 608\n\nTranscript\n```\nWEBVTT\n\n00:24.040 --> 00:29.120\n<Speaker 3>What we liked about cognitive services offerings were that they had a much higher fidelity.\n\n00:29.600 --> 00:32.880\n<Speaker 3>And they sounded a lot more like an actual human voice.\n\n00:33.680 --> 00:37.200\n<Speaker 4>Orlando ground 9555 requesting the end of pushback.\n\n00:38.680 --> 00:41.280\n<Speaker 4>9555 request to end pushback received.\n```\n\nKey Frames\n- 00:23.100 ![](keyFrame.23100.jpg)\n- 00:24.833 ![](keyFrame.24833.jpg)\n...\n- 00:42.633 ![](keyFrame.42633.jpg)\n- 00:43.233 ![](keyFrame.43233.jpg)",
        "fields": {
          "Summary": {
            "type": "string",
            "valueString": "The focus shifts back to an interview with another man discussing the high fidelity and human-like quality of voices produced by cognitive services offerings. The segment concludes with visuals of an airplane on the tarmac, ground crew directing the plane, and the plane preparing for pushback, accompanied by realistic audio communications between ground control and the aircraft."
          }
        },
        "kind": "audioVisual",
        "startTimeMs": 23100,
        "endTimeMs": 43233,
        "width": 1080,
        "height": 608,
        "KeyFrameTimesMs": [
          23100,
          24833
          /*... (19 additional keyframes)*/
        ],
        "transcriptPhrases": [
          {
            "speaker": "Speaker 3",
            "startTimeMs": 24040,
            "endTimeMs": 29120,
            "text": "What we liked about cognitive services offerings were that they had a much higher fidelity.",
            "confidence": 0.937,
            "words": [
              {
                "startTimeMs": 24040,
                "endTimeMs": 24240,
                "text": "What"
              },
              {
                "startTimeMs": 24240,
                "endTimeMs": 24320,
                "text": "we"
              }
              /*... (13 additional words)*/
            ],
            "locale": "en-US"
          },
          {
            "speaker": "Speaker 3",
            "startTimeMs": 29600,
            "endTimeMs": 32880,
            "text": "And they sounded a lot more like an actual human voice.",
            "confidence": 0.823,
            "words": [
              {
                "startTimeMs": 29600,
                "endTimeMs": 30080,
                "text": "And"
              },
              {
                "startTimeMs": 30080,
                "endTimeMs": 30160,
                "text": "they"
              }
              /*... (9 additional words)*/
            ],
            "locale": "en-US"
          },
          {
            "speaker": "Speaker 4",
            "startTimeMs": 33680,
            "endTimeMs": 37200,
            "text": "Orlando ground 9555 requesting the end of pushback.",
            "confidence": 0.823,
            "words": [
              {
                "startTimeMs": 33680,
                "endTimeMs": 34160,
                "text": "Orlando"
              },
              {
                "startTimeMs": 34160,
                "endTimeMs": 34600,
                "text": "ground"
              }
              /*... (6 additional words)*/
            ],
            "locale": "en-US"
          },
          {
            "speaker": "Speaker 4",
            "startTimeMs": 38680,
            "endTimeMs": 41280,
            "text": "9555 request to end pushback received.",
            "confidence": 0.823,
            "words": [
              {
                "startTimeMs": 38680,
                "endTimeMs": 39600,
                "text": "9555"
              },
              {
                "startTimeMs": 39600,
                "endTimeMs": 40080,
                "text": "request"
              }
              /*... (4 additional words)*/
            ],
            "locale": "en-US"
          }
        ],
        "cameraShotTimesMs": [
          1467,
          3233
          /*... (13 additional camera shots)*/
        ]
      }
    ]
  },
  "usage": {
    "videoHours": 0.013,
    "contextualizationTokens": 12222,
    "tokens": {
      "gpt-4.1-mini-input": 8976,
      "gpt-4.1-mini-output": 439
    }
  }
}

Nächster Schritt

Nachdem Sie nun wissen, wie Sie den Analysevorgang aufrufen, erfahren Sie mehr über das Erstellen von benutzerdefinierten Analysegeräten für Ihren Anwendungsfall.

Feedback

War diese Seite hilfreich?

Last updated on 2025-12-19