Get a Verbatim (Include Filler Words) Transcription from "client.begin_analyze"

[ahotrod]

Here's my working Audio Content Understanding - Content Extraction Python code after initialization & credentials:

client = AzureContentUnderstandingClient(
    endpoint    = AZURE_AI_ENDPOINT,
    api_version = AZURE_AI_API_VERSION,
    subscription_key = AZURE_AI_API_KEY,
    x_ms_useragent = "azure-ai-content-understanding-python/content_extraction",.
)

ANALYZER_AUDIO_FILE = '../data/presentation.m4a'
ANALYZER_ID = 'prebuilt-audioAnalyzer'

response = client.begin_analyze(ANALYZER_ID, file_location = ANALYZER_AUDIO_FILE)
result_json = client.poll_result(response, timeout_seconds = 600,
                                 polling_interval_seconds = 10)

I need a verbatim (with filler words) transcript which includes diarization and word timestamps to be returned. I see "SDK" suggestions for azure.ai.contentunderstanding to "include" the following arguments to "client.begin_analyze":

from azure.ai.contentunderstanding import ContentUnderstandingClient

response = client.begin_analyze(
    ANALYZER_ID, file_location = ANALYZER_AUDIO_FILE,
    mode = "verbatim",
    include_filler_words = True,
    include_speaker_diarization = True,
    include_word_timestamps = True
)

As far as I've found, the Azure Content Understanding SDK is still a Work-In-Process. Is this something that can be readily added to the existing headers of the "begin_analyze" method in this repo's content_understanding_client.py file?

Or something like the following "analyzer.verbatim.json" file:

{
  "$schema": "https://cognitiveservices.azure.com/schemas/2024-03-01/analyzer.json",
  "displayName": "Verbatim Transcription Analyzer",
  "description": "Captures exact speech with fillers, diarization, and word timestamps",
  "kind": "speechContentRecognizer",
  "properties": {
    "mode": "verbatim",
    "includeFillerWords": true,
    "includeSpeakerDiarization": true,
    "includeWordTimestamps": true
  }
}

Thanks in advance for any discussion / comments.

[ahotrod]

Custom "verbatim-tx-audioAnalyzer" analyzer successfully created and deployed. The base "prebuilt-audioAnalyzer" with a supposedly altered runtime configuration per the properties key:

verbatim_tx = {
    "$schema": "https://aka.ms/azure-content-understanding/analyzer-schema/v1",
    "kind": "speechContentRecognizer",
    "version": "2025-05-01-preview",
    "displayName": "verbatim-tx-audioAnalyzer",
    "description": "Verbatim transcription using prebuilt-audioAnalyzer to include fillers, diarization & timestamps",
    "baseAnalyzerId": "prebuilt-audioAnalyzer",
    "properties": {
        "mode": "verbatim",
        "includeFillerWords": True,
        "includeSpeakerDiarization": True,
        "includeWordTimestamps": True
    }
}

Calling "client.begin_analyze" with this analyzer's ID does not provide verbatim transcription. Furthermore, changing properties "includeSpeakerDiarization" & "includeWordStamps" to "False" is ignored. Seems all configuration properties are ignored.

Copilot suggests:
"Filler word inclusion is only supported in certain regions and may require preview access. Make sure:

• You're deploying in a supported region like West US or East US.
• Your Azure subscription has access to the verbatim mode and filler word support, which may be gated behind preview features."

Deployed in West US region.

How to find if our Azure subscription has access to the verbatim mode and filler word support? Does it even exist?

Is there a formal Azure Content Understanding Discussion group?

[ahotrod]

Creating & deploying this analyzer schema file:

{
  "kind": "speechContentRecognizer",
  "version": "2025-05-01-preview",
  "displayName": "verbatim-tx-audioAnalyzer",
  "description": "Verbatim transcription using prebuilt-audioAnalyzer to include fillers, diarization & timestamps",
  "baseAnalyzerId": "prebuilt-audioAnalyzer",
  "config": {
    "returnDetails": true,
    "disableContentFiltering": true
  },
  "properties": {
    "mode": "verbatim",
    "includeFillerWords": true,
    "includeSpeakerDiarization": true,
    "includeWordTimestamps": true,
    "enablePunctuation": true
  }
}

results in this analyzer:

{
  "analyzerId": "verbatim-tx-17c12049-c2f1-4439-8f5b-503bfddxxxxx",
  "description": "Verbatim transcription using prebuilt-audioAnalyzer to include fillers, diarization & timestamps",
  "createdAt": "2025-08-22T17:16:29Z",
  "lastModifiedAt": "2025-08-22T17:16:29Z",
  "baseAnalyzerId": "prebuilt-audioAnalyzer",
  "config": {
    "returnDetails": true,
    "disableContentFiltering": false
  },
  "warnings": [],
  "status": "ready",
  "processingLocation": "geography",
  "mode": "standard"
}

According to Copilot:
"As of August 2025, 2025-05-01-preview is the latest available API version for Azure AI Content Understanding’s audio analyzers, including prebuilt-audioAnalyzer. This version introduces expanded support for:

• Speaker diarization
• Word-level timestamps
• Filler word inclusion
• Custom analyzer templates
• Multilingual transcription across supported locales"

Any ideas? Any leads for further investigation? My project urgently needs this capability.