How to Get a Full Transcript from Teams Meetings

Tested prompts for transcribe microsoft teams meeting recording compared across 5 leading AI models.

BEST BY JUDGE SCORE Claude Opus 4.7 8/10

You recorded a Microsoft Teams meeting and now you need a readable, searchable transcript. Maybe the built-in Teams transcription was turned off, the recording was saved to OneDrive or SharePoint without captions, or you need a cleaner, more accurate version than what Teams auto-generated. Whatever the gap is, you have a video or audio file and you need text out of it.

The standard workflow is to extract the audio from your Teams recording, pass it through an AI transcription model, then clean up the output with a language model that can format speaker turns, fix proper nouns, and add punctuation. The prompt and model comparisons on this page cover exactly that pipeline.

This matters because raw Teams transcripts are often riddled with missed words, run-on sentences, and misattributed speakers. A properly processed transcript is something you can share with stakeholders, drop into a meeting summary tool, search for a specific decision, or feed into a CRM note. Getting the transcription step right determines the quality of everything downstream.

When to use this

This approach works best when you have a completed Teams meeting recording saved as an MP4 or audio file and need an accurate, structured transcript. It fits situations where the built-in Teams captions were not enabled, the auto-generated transcript is too noisy to use directly, or you need the transcript in a specific format for a downstream workflow.

The Teams meeting was recorded but transcription was not enabled before the call started
The auto-generated VTT or DOCX transcript from Teams has significant errors in technical terms, names, or product names
You need a clean speaker-labeled transcript to hand off for meeting minutes or legal documentation
You want to extract action items or decisions from a long recording without watching the whole video
You are processing a batch of archived Teams recordings for a compliance or knowledge-base project

When this format breaks down

The meeting is still in progress or was just recorded and the built-in Teams transcription is already running correctly. Wait for the native output before adding extra steps.
Your recording contains heavy crosstalk, multiple simultaneous speakers, or very poor audio quality. AI transcription accuracy drops significantly and the output will require more manual correction than it saves.
The recording is subject to strict data-residency or legal hold requirements that prohibit sending audio to a third-party API. Verify compliance before uploading to any external transcription service.
You only need a rough two-sentence summary of the meeting. Full transcription is overkill. Use Teams' built-in AI meeting recap if it is available in your license.

The prompt we tested

You are an expert meeting transcriptionist specializing in Microsoft Teams recordings. The user will describe their Teams meeting recording situation, and your job is to walk them through exactly how to obtain a complete, accurate transcript.

Follow these formatting and content rules:
Respond with a numbered step-by-step guide using short headers and concise bullet points under each step. Cover at least two methods (native Teams/Stream transcription and a third-party alternative like Otter.ai or Whisper), include export format options (DOCX, VTT, SRT, TXT), and keep the total response under 400 words. End with a brief 'Tips for Better Accuracy' section with 3 bullet points.

User's situation:
I have a 90-minute Microsoft Teams meeting recording from last week saved to OneDrive. It's a client strategy session with 6 participants, and I need a full transcript with speaker names and timestamps so I can share it with my team and pull action items. I didn't turn on live transcription during the call.

Provide a clear, step-by-step guide tailored to their scenario, including the best transcription method (built-in Teams transcription, Stream, or third-party tools), how to access or generate the transcript, how to edit and export it, and any tips for improving accuracy (speaker labels, timestamps, punctuation).

Model comparison

Model	Quality	Cost	Latency	Tokens
Claude Opus 4.7	8	$0.0805	17.5s	436 / 986
Claude Haiku 4.5	7	$0.0037	8.4s	305 / 670
GPT-5.4	7	$0.0113	15.7s	265 / 707
Gemini 2.5 Pro	3	$0.0020	20.8s	279 / 323
Grok 4.1 Fast Reasoning	6	$0.0002	11.3s	413 / 300

Outputs

Claude Opus 4.7 8/10

# Transcribing Your 90-Minute Teams Client Strategy Session

Since live transcription wasn't enabled, you'll generate one from the saved recording. Here are two reliable methods.

## Method 1: Native Microsoft Stream (Recommended)

### Step 1: Locate the Recording
- Open OneDrive → navigate to the **Recordings** folder
- Click the MP4 file to launch it in Stream (web player)

### Step 2: Generate the Transcript
- In the video player, click the **Video settings** (gear icon) on the right
- Select **Transcript and captions** → **Generate**
- Choose the spoken language (e.g., English)
- Wait 10–20 minutes for a 90-minute file to process

### Step 3: Review Speaker Labels
- Stream auto-detects speakers but may label them "Speaker 1, Speaker 2"
- Click any transcript line → rename speakers to match your 6 participants
- Edit misheard words inline by clicking the pencil icon

### Step 4: Export the Transcript
- Click the **⋯** menu above the transcript pane
- Choose **Download** → select format:
  - **VTT** – includes timestamps + speakers (best for sharing)
  - **DOCX** – paste into Word for clean formatting
  - Copy/paste to **TXT** for a simple version

## Method 2: Otter.ai (Better Speaker Diarization)

### Step 1: Export Audio from OneDrive
- Download the MP4 → optionally convert to MP3 with a free tool (HandBrake, VLC)

### Step 2: Upload to Otter
- Sign in at otter.ai → click **Import**
- Upload the file (Business plan handles 90 min easily)

### Step 3: Tag Speakers
- Otter separates voices automatically
- Assign names once per speaker; it propagates through the transcript

### Step 4: Export
- Click **Export** → choose **DOCX, TXT, SRT, or PDF**
- Include timestamps and speaker labels in export options

## Method 3: Whisper (Highest Accuracy, Free)
- Use OpenAI Whisper (large-v3 model) locally or via MacWhisper/Whisper WebUI
- Exports **SRT, VTT, TXT** with timestamps
- Pair with **pyannote.audio** for speaker diarization

---

## Tips for Better Accuracy

- **Clean the audio first**: Run the MP4 through Adobe Podcast Enhance (free) to reduce background noise before transcribing — critical for 6-person meetings with crosstalk.
- **Provide a name/term list**: In Otter or Whisper, pre-load client names, product terms, and acronyms so they aren't misspelled throughout.
- **Spot-check high-value moments**: Review the first 2 minutes and any section where action items are discussed — punctuation errors here distort meaning most when pulling takeaways.

Claude Haiku 4.5 7/10

# Getting a Transcript for Your Teams Meeting Recording

## Method 1: Microsoft Stream (Native Solution)

1. **Upload to Stream**
   - Open your OneDrive recording file
   - Right-click → Share → Copy link
   - Go to Stream.microsoft.com and upload via "Create" > "Upload video"
   - Allow 10–15 minutes for processing (longer for 90-minute files)

2. **Access Auto-Generated Transcript**
   - Once processing completes, click the video
   - Select "Transcript" tab on the right panel
   - Stream auto-generates captions with timestamps but **limited speaker identification**

3. **Export the Transcript**
   - Click three-dot menu → "Download transcript"
   - Available formats: **VTT, SRT, or TXT**
   - Copy-paste into Word for DOCX format if needed

**Limitation:** Speaker names aren't automatically labeled; you'll manually edit to add them.

---

## Method 2: Otter.ai (Third-Party Alternative)

1. **Upload Recording**
   - Create free Otter.ai account (300 minutes/month included)
   - Click "New conversation" → Upload your Teams recording file
   - Processing takes 15–30 minutes for 90 minutes of audio

2. **Review Transcript**
   - Otter provides **speaker labels automatically** (Speaker 1, 2, 3, etc.)
   - Edit speaker names directly in the interface
   - Timestamps are included by default
   - Accuracy typically 85–95% depending on audio clarity

3. **Export Options**
   - Download as **DOCX, PDF, or TXT**
   - Copy-paste transcript directly into your team doc
   - Share link with team for collaborative editing

---

## Method 3: OpenAI Whisper (Free, Advanced Users)

- Upload file to Whisper API or desktop app
- Generates highly accurate transcripts with timestamps
- Requires minimal technical setup; export as **TXT, VTT, or JSON**

---

## Quick Comparison

| Feature | Stream | Otter.ai | Whisper |
|---------|--------|----------|---------|
| Cost | Free | Free (300 min/mo) | Free |
| Speaker Labels | Manual | Automatic | Manual |
| Accuracy | 80–85% | 85–95% | 90–98% |
| Export Formats | VTT, SRT, TXT | DOCX, PDF, TXT | TXT, VTT, JSON |

---

## Tips for Better Accuracy

- **Clear Audio:** Position speakers close to microphones; minimize background noise for best results
- **Edit Speaker Names:** After auto-generation, manually replace "Speaker 1" with actual names for clarity when sharing
- **Timestamp Review:** Check timestamps align with key discussion points; use them to reference action items easily

GPT-5.4 7/10

1. **Best option for your recording**
- Since the meeting is already saved to **OneDrive** and live transcription was not enabled, the best first method is **Microsoft Stream (on SharePoint/OneDrive)**.
- This can generate a transcript **after the meeting** from the recording file.
- Best for keeping data inside Microsoft 365 and preserving **timestamps**.

2. **Method 1: Generate transcript in Stream**
- Go to **OneDrive** and locate the Teams recording.
- Open the video in **Stream**.
- In the video page, look for **Video settings** or **Transcript and captions**.
- Choose **Generate transcript** or **Autogenerate captions**.
- Select the correct **spoken language**.
- Wait for processing; a 90-minute file may take a while.

3. **Review speaker names and timestamps**
- Open the transcript pane beside the video.
- Stream usually includes **timestamps** automatically.
- Speaker identification may be limited unless Teams/meeting metadata is available; review manually for the 6 participants.
- Edit obvious errors:
- client names
- strategy terms
- acronyms
- action items
- punctuation

4. **Export and share**
- In Stream, look for **Download transcript/captions**.
- Common export options:
- **VTT**: best for timestamps/captions
- **SRT**: subtitle-friendly
- **TXT**: simple text
- If you need a shareable meeting summary:
- open the file in Word
- paste transcript
- save as **DOCX**
- Use the transcript to extract action items for your team.

5. **Method 2: Check Teams meeting details**
- Open the meeting in **Teams Calendar** or **Chat**.
- If Teams created any post-meeting artifacts, check the **Recap** tab.
- If a transcript is unexpectedly available, you may be able to:
- view it
- edit speaker names
- download it
- Export options depend on your tenant, but **VTT** is most common.

6. **Method 3: Third-party for better speaker labeling**
- If Stream does not label all 6 speakers clearly, use **Otter.ai** or **Whisper-based tools**.
- Download the MP4 from OneDrive, then upload it.
- Otter.ai:
- good for **speaker separation**
- editable transcript
- export to **TXT, DOCX, SRT**
- Whisper tools:
- often very accurate
- can produce **TXT, SRT, VTT**
- speaker names may require manual assignment unless diarization is included

7. **Recommended workflow**
- Try **Stream first** for secure Microsoft-native transcription.
- If speaker names are weak, run the same file through **Otter.ai or Whisper**.
- Use Stream for timestamps, then merge with third-party speaker labeling in **Word DOCX** for team sharing.

**Tips for Better Accuracy**
- Review and replace generic labels like “Speaker 1” with actual participant names.
- Edit punctuation and formatting so action items and decisions are easy to scan.
- If possible, use the participant list/chat to verify who was speaking at key timestamps.

Gemini 2.5 Pro 3/10

Of course. Since you didn't enable live transcription during the meeting, you can easily generate it after the fact. Here is a step-by-step guide to get the complete, speaker-identified transcript you need.

### 1. Generate the Transcript in Teams/Stream

This method uses Microsoft's built-in tools and is the fastest way to start.

*   **Locate the Recording:** Find the meeting in your Teams calendar or chat history. Click the recording to open it in a new browser window (this is the Microsoft Stream/SharePoint player).
*   **Generate Transcript:** In the video player window, find the **Video settings** panel on the right. Select **Transcript and captions** and click **Generate**.
*   **Select Language:** You will be prompted to choose the language spoken in the meeting. Select English and confirm. The generation may take some time for a 90-minute file.

### 2. Review and Export the Native Transcript

Once the transcript is ready, you'll need to clean it up and export it.

*   **Edit Speaker Names:** The transcript will appear in a panel next to the video. It will label speakers generically (e.g., "Speaker 1," "Speaker 2"). Hover over a speaker's name, click the pencil icon, and type their actual name. Choose "Change all" to update it throughout the transcript.
*   **Correct Text:** Review the text for errors and make corrections directly in the transcript panel.
*   **Download:** Above the

Grok 4.1 Fast Reasoning 6/10

1. **Access Recording in Teams (Native Method)**
   - Open Teams app/web > Go to chat with recording > Click recording file.
   - Play recording > Click "Transcript" or "..." > "Generate transcript" (processes ~90-min file in 10-20 min; auto-detects 6 speakers).

2. **Edit & Export Native Transcript**
   - In transcript pane: Edit speaker names/labels, timestamps, punctuation.
   - Export: "..." > Download as TXT, VTT, SRT, or copy to Word for DOCX.

3. **Alternative: Upload to Otter.ai (Third-Party)**
   - Sign up at otter.ai (free tier for 90 min) > New Otter > Import file > Upload MP4 from OneDrive.
   - Auto-generates transcript with speaker ID, timestamps > Edit names for 6 participants.

4. **Edit & Export Otter Transcript**
   - Review/edit in app > Export as DOCX, TXT, SRT, VTT > Share link or download.

5. **Share & Extract Action Items**
   - Upload edited DOCX to OneDrive/SharePoint > Use Teams/Word "Copilot" or search for "[Action]" keywords.

**Tips for Better Accuracy**
- Assign unique speaker names pre-transcription if possible (Teams/Otter settings).
- Ensure clear audio; minimize accents/background noise in future recordings.
- Cross-verify with playback at timestamps for multi-speaker overlaps.

(248 words)

What makes these work

01

Extract audio before transcribing
Teams recordings save as MP4 video files. Passing just the audio track to a transcription model is faster, cheaper per token, and reduces noise from processing video frames. Use a tool like FFmpeg to strip the audio to a WAV or MP3 before sending to the transcription API.
02

Provide speaker names in your prompt
AI transcription models do better speaker attribution when you tell them who was on the call. Include a line like 'Speakers on this call: Sarah (Product Manager), Mark (Engineer)' at the start of your prompt. This dramatically reduces misattribution in multi-speaker recordings.
03

Specify your output format explicitly
The model will not know you want a CRM note, a bullet-point action list, or a timestamped transcript unless you say so. Add a format instruction to every prompt: 'Format the output as a timestamped transcript followed by a numbered action item list.' Vague prompts produce vague output.
04

Chunk long recordings into segments
Most Teams meetings run 30 to 60 minutes. Sending the entire audio at once can exceed context limits and hurt accuracy late in the transcript. Break recordings into 10-15 minute chunks, transcribe each, then pass the combined raw text through a second prompt to produce the final summary or action list.

More example scenarios

#01 · Product team retrospective needing action items

Input

Transcribe the following Teams meeting audio excerpt and format it with speaker labels and timestamps. Then list all action items mentioned. Audio content: 'Sarah: We need to fix the onboarding flow before launch. Mark: I can have a prototype by Thursday. Sarah: Great, let us sync Friday morning. Also someone needs to update the help docs. Mark: I will assign that to Jamie.'

Expected output

Transcript:
[00:00] Sarah: We need to fix the onboarding flow before launch.
[00:04] Mark: I can have a prototype by Thursday.
[00:07] Sarah: Great, let us sync Friday morning. Also someone needs to update the help docs.
[00:12] Mark: I will assign that to Jamie.

Action Items:
- Mark to deliver onboarding flow prototype by Thursday
- Sarah and Mark to sync Friday morning
- Jamie to update the help documentation

#02 · Legal team recording a client intake call

Input

Transcribe this Teams recording excerpt with accurate speaker attribution and flag any statements that sound like commitments or agreements. Speakers are an attorney and a new client discussing case scope. Content: 'Attorney: We can take this on a contingency basis. Client: So I pay nothing upfront? Attorney: Correct, our fee is 33 percent of any settlement. Client: I agree to those terms.'

Expected output

Transcript:
Attorney: We can take this on a contingency basis.
Client: So I pay nothing upfront?
Attorney: Correct, our fee is 33 percent of any settlement.
Client: I agree to those terms.

Flagged Commitments:
- Attorney commits to contingency representation
- Fee structure stated as 33 percent of settlement
- Client verbally agrees to stated terms

#03 · Sales team call for CRM note entry

Input

Transcribe this 60-second Teams sales call recording and produce a CRM-ready summary. Speakers: account executive and prospect. Key topics discussed: pricing, Q3 implementation timeline, and a follow-up demo request. Format output as a brief CRM note under 100 words.

Expected output

CRM Note - Call Date: [date]
Contact: Prospect (name TBC)
AE: [Account Executive Name]

Prospect expressed interest in enterprise tier pricing. Budget approval expected by end of Q2. Requested implementation to begin Q3. Asked for a technical demo focused on API integration capabilities. AE to send pricing sheet and schedule demo within 48 hours. Deal stage: Discovery. Next step: Demo call.

#04 · HR conducting a recorded interview debrief

Input

Transcribe the following Teams interview debrief recording between two hiring managers discussing a candidate. Preserve exact quotes where either manager makes an evaluative statement. Flag any language that could be considered legally problematic in a hiring context.

Expected output

Transcript excerpt:
Hiring Manager 1: Her technical answers were strong, especially on system design.
Hiring Manager 2: Agreed. I did wonder about her availability given she mentioned having young kids.

Flagged Language:
- "availability given she mentioned having young kids" -- Reference to family status in a hiring decision context may constitute discrimination under employment law. Recommend removing this from official documentation and focusing feedback on job-relevant criteria only.

#05 · Engineering team extracting decisions from a long architecture review

Input

Transcribe this 45-minute Teams architecture review recording and produce only the final decisions made, not the full discussion. Format as a numbered decision log with the decision owner's name and any stated rationale. The team discussed database choice, API versioning strategy, and deployment pipeline.

Expected output

Decision Log:
1. Database: PostgreSQL selected over MongoDB. Owner: Lead Architect. Rationale: Existing team expertise and relational data model fit.
2. API Versioning: URI versioning adopted (e.g., /v1/, /v2/). Owner: Backend Lead. Rationale: Simpler for external consumers to manage.
3. Deployment Pipeline: GitHub Actions chosen over Jenkins. Owner: DevOps Lead. Rationale: Lower maintenance overhead and native GitHub integration.

Common mistakes to avoid

Using the raw VTT file without cleaning it
Teams exports transcripts as VTT caption files that contain timing codes, repeated fragments, and filler words at high density. Pasting a raw VTT directly into a language model prompt wastes tokens on formatting noise and degrades output quality. Strip the timing codes and clean line breaks first.
Skipping speaker identification setup
Sending audio without telling the model who the speakers are produces transcripts labeled 'Speaker 1' and 'Speaker 2' throughout. These labels are useless for meeting notes or CRM entries. Always provide a speaker roster in the prompt, even if it is just first names.
Not verifying proper nouns and product names
AI transcription models frequently mishear company names, product names, acronyms, and technical terms. A model might transcribe 'Salesforce' as 'sales force' or an internal product name as something phonetically similar. Always do a quick scan of the transcript for named entities before distributing it.
Assuming one model handles both transcription and summarization equally well
Transcription (audio to text) and summarization (text to condensed text) are different tasks that often benefit from different models or two separate prompt steps. Running a single prompt that asks a language model to both transcribe and summarize in one pass often produces a compressed version rather than a faithful transcript.
Uploading sensitive recordings to public APIs without checking policy
Teams recordings from HR, legal, finance, or executive meetings may contain regulated or confidential information. Many organizations have policies prohibiting upload of meeting recordings to external AI services. Confirm your data handling policy before using any cloud-based transcription API on internal recordings.

Related queries

Frequently asked questions

Can I get a transcript from a Teams recording if transcription was not turned on during the meeting?

Yes. Download the meeting recording from OneDrive or SharePoint as an MP4 file. Extract the audio track and run it through an AI transcription service such as Whisper, Azure AI Speech, or a similar API. The resulting transcript will be comparable in accuracy to what live transcription would have produced, sometimes better on cleaned audio.

Where does Microsoft Teams save meeting recordings?

For most Microsoft 365 tenants, Teams recordings are automatically saved to the OneDrive of the person who started the recording, inside a folder called 'Recordings.' In channel meetings, they save to the SharePoint document library of that team's channel. You can find the file by opening the meeting chat and clicking the recording link, then navigating to its file location.

How accurate is the built-in Teams transcription compared to AI tools?

Microsoft Teams native transcription uses Azure AI Speech and is generally accurate for clear speech in English, but it struggles with heavy accents, technical jargon, and poor microphone audio. Third-party tools built on Whisper large or similar models often perform better on domain-specific vocabulary because you can provide a custom vocabulary or prompt context before transcribing.

How do I download a Teams meeting transcript as a Word document?

During or after a meeting, open the meeting in Teams, go to the Recap or Transcript tab, and click the download button. Teams offers a DOCX download option that formats the transcript with speaker names and timestamps. This option is only available if transcription was enabled before or during the meeting by a meeting organizer with the right license.

What is the best way to transcribe a long Teams recording, like a two-hour all-hands meeting?

Split the recording into segments of 10 to 15 minutes using a tool like FFmpeg before transcribing. Transcribe each segment separately to maintain accuracy and stay within API context limits. After transcription, concatenate the text segments and run a single summarization prompt over the full combined transcript to produce action items or a meeting summary.

Does Microsoft Copilot in Teams automatically transcribe and summarize meetings?

Yes, if your organization has a Microsoft 365 Copilot license, Copilot can generate meeting summaries, transcripts, and action items automatically after a recorded meeting. However, this requires the meeting to have had transcription enabled during the session and requires the Copilot add-on license, which is separate from standard Microsoft 365 plans.