Documentation question: Does Microsoft Stream transcription on SharePoint use Azure AI Speech?

Question

Documentation question: Does Microsoft Stream transcription on SharePoint use Azure AI Speech?

Catherine M. Stockdale 0

Hello,

I'm working on a research project that requires me to accurately describe the technology stack used for transcription within Microsoft 365, and I'm hoping you can help me clarify a technical question that I have not been able to confirm through public documentation.

My specific questions are:

Does the transcription functionality in Microsoft Stream — which generates transcripts and closed captions for videos uploaded to SharePoint — use Azure AI Speech as its underlying automatic speech recognition (ASR) engine?
Is there an existing Microsoft document or technical reference that describes how transcripts are generated for videos uploaded to SharePoint, including the underlying technology stack?

I have found Microsoft documentation confirming that:

Azure AI Speech powers transcription scenarios in Microsoft Teams (captioning), Microsoft 365 (dictation), and Microsoft Edge (Read Aloud) (Microsoft Learn — What Is Azure Speech?).
Microsoft Stream automatically generates transcripts and WebVTT (.vtt) caption files for videos uploaded to SharePoint (Microsoft Learn — Updates to transcript functionality in Microsoft Stream).

However, I have not been able to find documentation explicitly confirming that SharePoint uses Azure AI Speech as its ASR engine, nor have I found a comprehensive reference describing the technology behind SharePoint's transcript generation. If you can confirm this connection or point me to relevant documentation, I would greatly appreciate it. If this information is not publicly available, a brief note to that effect would also be helpful.

Thank you for your time.

0 comments

2 answers

Your answer

Answer 1

SRILAKSHMI C 18,990 Microsoft External Staff Moderator

Hello @Catherine M. Stockdale

Thanks for your detailed question

Does Stream (on SharePoint) use Azure AI Speech?

From a platform perspective, yes Microsoft Stream (on SharePoint) transcription is powered by Microsoft’s cloud-based speech recognition services, which are part of the same underlying platform as Azure AI Speech.

This is the same core “Speech-to-Text” technology that powers experiences across:

Microsoft Teams (live captions and transcription)
Microsoft 365 (dictation in Word/Office)
Microsoft Edge (Read Aloud)

Additionally:

Stream does not use a custom on-premises ASR engine

It relies on Microsoft-hosted speech services running in Azure for automatic transcription and caption generation

Documentation

While the underlying technology is shared:

Microsoft does not explicitly document that Microsoft Stream directly consumes Azure AI Speech as a customer-facing API or dependency.

Instead, Stream uses Microsoft’s internal speech services, which are built on the same speech platform that powers Azure AI Speech. These internal services may differ from the publicly exposed APIs in terms of implementation and integration.

Is there a detailed architecture or pipeline document?

At present, there is no publicly available deep-dive or end-to-end architecture document specifically for Stream transcription in SharePoint.

What is available:

Documentation confirming that Microsoft Stream:

Automatically generates transcripts
Produces WebVTT (.vtt) caption files

The “Updates to transcript functionality in Microsoft Stream” article (functional overview)

General documentation for Azure AI Speech.

What is not available:

Detailed transcription pipeline for Stream

Architecture diagrams for SharePoint video processing

Explicit confirmation of direct Azure AI Speech API usage

Please refer this

Updates to transcript functionality in Microsoft Stream https://learn.microsoft.com/stream/transcript-functionality-updates

What Is Azure Speech? (Azure Cognitive Services Speech Service overview) https://learn.microsoft.com/azure/ai-services/speech-service/overviewI Hope this helps. Do let me know if you have any further queries.

If this answers your query, please do click Accept Answer and Yes for was this answer helpful.

Thank you!

Catherine M. Stockdale 0 Reputation points

2026-05-05T18:30:41.49+00:00

Thank you for the detailed responses to my earlier questions. I have two follow-up questions if you have time.

Question 1: DAX Copilot (Nuance Dragon Ambient eXperience) ASR backend

Does Microsoft publicly document the speech recognition technology used by DAX Copilot (Microsoft Nuance Dragon Ambient eXperience for ambient clinical documentation)? Specifically, is there technical documentation that confirms whether DAX Copilot directly consumes Azure AI Speech as a customer-facing API or dependency, or is the underlying ASR backend abstracted in the same way it is for Microsoft Stream on SharePoint?

Question 2: Scope of "profanity" content moderation in SharePoint transcription

Does Microsoft publicly document what categories of language are included in the default profanity masking applied to SharePoint transcripts? Specifically, does "profanity" in this context refer narrowly to swear words and slurs, or does the moderation also apply to other categories of language — for example, anatomical terminology (such as words referring to genitalia or other body parts), sexual health terminology, or other clinically relevant terms that might be flagged by content moderation classifiers?

If detailed documentation about either of these is not publicly available, a brief confirmation of that would also be helpful. Thank you again for your help!
Catherine M. Stockdale 0 Reputation points

2026-05-05T18:33:01.3833333+00:00

Thank you for the answer to my question above. I have one more question:

Does Microsoft publicly document the speech recognition technology used by DAX Copilot (Microsoft Nuance Dragon Ambient eXperience for ambient clinical documentation)? Specifically, is there technical documentation that confirms whether DAX Copilot directly consumes Azure AI Speech as a customer-facing API or dependency, or is the underlying ASR backend abstracted in the same way it is for Microsoft Stream on SharePoint?
SRILAKSHMI C 18,990 Reputation points Microsoft External Staff Moderator

2026-05-06T16:57:28.15+00:00
Hi Catherine M. Stockdale,

Thank you for your follow-up questions

Question 1: DAX Copilot (Nuance Dragon Ambient eXperience) – ASR backend

Regarding Nuance Dragon Ambient eXperience:

Publicly available information

DAX Copilot is built on Nuance Dragon Medical speech recognition technology, combined with ambient AI and clinical language understanding.

It is delivered as an end-to-end healthcare solution, rather than a set of exposed platform APIs.

At this time There is no public Microsoft documentation that explicitly confirms:

Direct use of Azure AI Speech as a customer-facing API or dependency, or

A detailed architecture of the ASR pipeline used by DAX Copilot

Clarification

The ASR capability in DAX Copilot is abstracted within Nuance’s Dragon Medical platform

While Microsoft and Nuance technologies are integrated at a broader platform level, the underlying speech services are not exposed or documented as Azure AI Speech usage

This is similar in principle to other Microsoft 365 experiences where underlying services are not surfaced as direct API dependencies

Question 2: Profanity masking in SharePoint transcription

Publicly available information

Microsoft Stream (on SharePoint) provides automatic transcript generation with profanity filtering/masking

This feature is designed to improve content appropriateness in generated captions

Currently Microsoft does not publicly document a detailed definition or taxonomy of what is included in “profanity” for SharePoint transcription

Specifically, there is no official documentation that:

Lists exact words or categories

Defines whether filtering includes:

Anatomical or medical terminology

Sexual health terms

Other context-sensitive language

Clarification

In general, “profanity” refers to offensive language such as swear words or slurs

The filtering behavior is:

Model-driven and context-aware

Not based on a fixed, publicly documented dictionary

As a result, behavior may vary depending on context, and not all sensitive or clinical terms are necessarily masked

Thank you!
SRILAKSHMI C 18,990 Reputation points Microsoft External Staff Moderator

2026-05-07T10:47:30.2933333+00:00

Hi Catherine M. Stockdale,

Following up to see if the above answer was helpful. If this answers your query, please do click Accept Answer and Yes for was this answer helpful. And, if you have any further query do let us know.

Thank you!

Answer 2

Hi ,

Thanks for reaching out to Microsoft Q&A.

Short answer: There is no public ms documentation that explicitly confirms that ms Stream (on SharePoint) transcription uses Azure AI Speech as its underlying ASR engine.

explanation: While Azure AI Speech is officially documented as the speech-to-text engine behind services like Teams captions, dictation, and Edge Read Aloud, Microsoft has not published a technical reference that directly ties SharePoint/Stream (on SharePoint) transcript generation to that same service. Stream documentation confirms automatic transcript and VTT generation, but the underlying implementation is abstracted and described only as part of Microsoft 365 intelligent services. In practice, it is highly likely that Azure AI Speech or a closely related internal variant is used, but from a research or documentation standpoint, you should state that the exact ASR backend for SharePoint video transcription is not publicly disclosed by Microsoft.

Please 'Upvote'(Thumbs-up) and 'Accept' as answer if the reply was helpful. This will be benefitting other community members who face the same issue.

Share via

Documentation question: Does Microsoft Stream transcription on SharePoint use Azure AI Speech?

2 answers

Your answer