Where does Microsoft Teams save recordings?

For organizational tenants: recordings save to Microsoft Stream (stream.office.com) for channel meetings, or to OneDrive's Recordings folder for non-channel meetings. The recording link appears in the Teams chat for the meeting. For personal/free accounts, recording is generally not available.

My Teams .mp4 is too large for the 25 MB upload — what now?

Extract just the audio: 'ffmpeg -i teams.mp4 -vn -c:a copy teams.m4a'. The audio-only is roughly 10× smaller. If still too large, split: 'ffmpeg -i teams.m4a -f segment -segment_time 1500 -c copy chunk%03d.m4a' — 25-min chunks.

Teams already shows a transcript — why use this?

Teams' built-in transcript is org-tenant only and tied to your Microsoft 365 license; it does not work on personal accounts, and the live transcript is sometimes not retained. ScribeForge is account-agnostic — drop any .mp4, get text — useful when you receive a Teams recording from outside your org or your tenant didn't enable transcription.

Is enterprise data safe? Are recordings stored?

Audio is sent to xAI's Grok STT API for processing, then deleted. ScribeForge retains neither audio nor transcript. The transcript exists only in your browser session. For tenant compliance review, see /privacy.

Will it identify each speaker on the call?

Yes — Grok STT does automatic speaker diarization (Speaker 1, Speaker 2, …). Rename to actual names afterward in the .txt. Accuracy depends on each participant having a few seconds of speech and a distinguishable voice.

Use case · Updated April 29, 2026

How to Transcribe a Microsoft Teams Recording — Free

Drop the .mp4 from Stream, SharePoint, or OneDrive — get a transcript with timestamps and speakers in seconds.

Yes — any Teams recording can be transcribed in browser. Download the .mp4 from Microsoft Stream, the channel's SharePoint folder, or your OneDrive Recordings folder, then drop it into ScribeForge. Grok STT returns text in 10–30 seconds with phrase-level timestamps and speaker labels.

Where Teams stores your recording

Meeting type	Storage location
Channel meeting	SharePoint Site for the channel (Recordings folder)
Non-channel meeting (org)	OneDrive of the organizer (Recordings folder)
Anything in Stream	stream.office.com — search by date/title
Personal account	Recording generally not available — needs Microsoft 365 license

The 3-step workflow

Download as .mp4. In Stream: open the recording → Download video. In SharePoint or OneDrive: three-dot menu → Download. Save anywhere local.
Drop into ScribeForge — go to scribeforge.tech, drag the file onto the upload zone. 25 MB cap fits ~8–10 minutes of default Teams .mp4.
Click Transcribe. Grok extracts the audio track and returns timestamped text with speakers in 10–30 seconds.

Long meetings — extract audio first

The Teams .mp4 includes 720p or 1080p video. For transcription only, extract the audio:

ffmpeg -i teams.mp4 -vn -c:a copy teams.m4a

Resulting .m4a is ~1/10 the size. If still over 25 MB, split into 25-minute chunks:

ffmpeg -i teams.m4a -f segment -segment_time 1500 -c copy chunk%03d.m4a

Transcribe each chunk separately. Concatenate the resulting .txt files in order.

"But Teams has a built-in transcript"

It does — when your tenant has it enabled and the meeting was set to record. Two real-world reasons people end up here:

Tenant didn't enable transcription for the meeting (admin policy or per-meeting setting).
You received the .mp4 from outside your org — a vendor, partner, candidate. No transcript came with it.
Personal Microsoft accounts don't have access to Teams transcription.
The auto-transcript got noisy — Grok STT consistently outperforms Microsoft's STT on diarization and entity recognition (per xAI's published benchmark, 5.0% vs 13.5%+).

Compliance + data handling

Audio is sent to xAI's Grok STT API and deleted immediately after transcription returns. ScribeForge does not retain the audio or the transcript — the text exists only in your browser tab. If your org requires that data never leave a specific region, get cleared with the xAI API documentation first; ScribeForge is a pass-through to that API.

Common questions

Can I transcribe a Teams recording from someone else's tenant?

Yes — ScribeForge is account-agnostic. As long as you can download the .mp4 (typically a "Download" link in the Stream or SharePoint share), you can transcribe it.

Are speakers identified by name?

Speakers are labeled "Speaker 1", "Speaker 2", etc. — diarization is voice-based, not identity-based. Rename to real names manually after transcription.

Does it work on Linux/macOS/Windows?

Yes — runs in any modern browser. ScribeForge is a web app, no install. ffmpeg is the only thing you might need locally for splitting long meetings.

How accurate is the transcript on conference-call audio?

Per xAI's official benchmark, Grok STT reports a 5.0% phone-call entity error rate — versus Deepgram 13.5% and AssemblyAI 21.3%. Real-world meeting transcripts (clean voice, mid-quality mic) typically come back at 7–10% WER. Full benchmark vs Whisper and Deepgram.

Drop a Teams .mp4 and read it in 30 seconds.

Transcribe Teams meeting free →

No account · No credit card · 2 free uses/day per IP