How to Transcribe In-Person Meetings Recorded on Your Phone with AI
Zoom meetings have automatic transcription. Teams does too. But what about the meeting that happens in a room, with people sitting around a table, no camera and no connected platform? That one still depends on someone with a notebook — or a memory that’s already a bit blurry the next day.
The problem isn’t a lack of technology. It’s that most tutorials focus on virtual meetings, as if in-person meetings don’t exist anymore. They do, and there are plenty of them. The difference is that nobody teaches you the full path: record with your phone, send the audio to an AI, and get back organized text.
That’s exactly what this guide does. From your phone’s voice recorder to a ready-to-share minutes document — no complicated setup, no audio lab, no hiring a transcriber.
Why In-Person Meetings Are Harder to Transcribe
In a virtual meeting, each participant has their own microphone. The audio comes in clean, separated, and the AI can identify who’s speaking without effort. Accuracy exceeds 95%.
In a room, it’s different. A single phone captures all voices at the same time, mixed with air conditioning noise, chairs scraping, and side conversations. Whoever is closest to the phone comes through clearly. Whoever is on the other side of the table becomes a distant mumble in the recording. Transcription accuracy drops to 70–85% if the recording is poorly done.
The good news: with the right care when recording and the proper transcription tool, you can get close to virtual meeting quality. Without investing in professional equipment.
Step 1: Record the Meeting with Your Phone
You don’t need a sophisticated app to record. The voice recorder that comes built into your phone does the job.
On iPhone
Open the Voice Memos app. Tap the red button to start recording. The file is saved in m4a format, which works well with most AI transcription services.
On Android
Use the native Voice Recorder or Google Recorder (if available on your device). Google Recorder has the advantage of transcribing in real time, but it only works well on select phone models.
Where to Place the Phone
This is the single most important factor for transcription quality:
- Center of the table, equidistant from all participants — not in your pocket, not at the end of the table facing only you
- Microphone facing the participants — on most phones, the microphone is on the bottom edge. Lay the phone on its side, not standing upright
- Stable surface — avoid glass or metal tables that resonate. Place a notebook or cloth underneath to dampen vibrations
- Away from the air conditioner and open windows
Before Recording: Three Quick Checks
- Do Not Disturb mode on — a notification mid-recording becomes a loud thud in the audio
- Bluetooth off — if your phone is paired with earbuds, audio may be redirected without you noticing
- Notify participants — tell them you’ll be recording. It’s not just courtesy; the LGPD (Brazil’s General Data Protection Law) requires consent to process voice as personal data
Step 2: Send the Audio for AI Transcription
This is where Sintesy comes in as the most direct path for anyone working in Brazilian Portuguese.
The difference between Sintesy and foreign services isn’t just the language. It’s that Sintesy was built for PT-BR from day one. This means it handles regional accents better, along with expressions from spoken Brazilian Portuguese and the verbal habits that show up in any real meeting — “né”, “tá”, “tipo assim”, “entende”.
The Workflow
- Meeting over? Open the recorder and export or share the audio file
- Send the audio to Sintesy — via the app or the web dashboard
- The AI processes the full audio and returns a text transcription, with paragraphs organized by conversation segment
- Along with the transcription, you receive an automatic summary with the key points and next steps identified in the conversation
No need to listen to the audio again. No need to pause transcription to correct it every minute. The process is upload and wait — in minutes, you have the text on screen.
Step 3: Review and Organize the Transcription
AI transcription is good, but not perfect. Especially in in-person meetings, some adjustments are expected:
What Usually Needs Correction
- Proper names: the AI may get names of people, companies, and products wrong if they aren’t common in its training data
- Technical terms: jargon from specific fields (medical, legal, engineering) may come out as a similar-sounding word
- Who said what: with phone audio, speaker identification is approximate. If the meeting has 4 or more people, review the attribution of statements
What Usually Doesn’t Need Correction
- Complete sentences and overall meaning: even with occasional errors, the context of the conversation comes through correctly
- Decisions and next steps: Sintesy’s automatic summary typically captures well what was decided and who was assigned what
- Chronology: the order of topics in the transcription follows the order of the conversation, without skipping sections
How to Review Without Wasting Time
Don’t read the entire transcription from start to finish. Do it this way instead:
- Read the summary first — if the main decisions are there, the transcription is probably good
- Search for people’s names — use Ctrl+F to find where each person was mentioned and check whether the attribution makes sense
- Check sensitive sections — if there was a financial, legal, or contractual decision, jump straight to that section and read carefully
- Adjust and done — the text is in editable format, just correct and save
Step 4: Turn the Transcription into Meeting Minutes
Having the transcription is useful, but what the team usually needs is meeting minutes — a structured document with date, participants, topics discussed, decisions, and next steps.
Sintesy already delivers an automatic summary with this basic structure. But if you want something more formal, you can take the raw transcription and ask the AI to reorganize it into a minutes format:
- Date and participants (fill in manually)
- Topics discussed (the AI groups by theme)
- Decisions made (the AI extracts from context)
- Next steps and responsible parties (the AI identifies when someone takes on a task)
What used to take 40 minutes of listening and typing now takes 5 minutes of review. The difference is that you’re reviewing something that’s already done instead of building from scratch.
Why Not Just Use the Phone’s Built-in Transcription?
The iPhone (iOS 17+) and Google Recorder on some Android devices already transcribe audio on the device itself. It seems sufficient. But there are limitations that show up when you use them day to day:
- iPhone’s native transcription: works well for short audio and in English. In Brazilian Portuguese, quality is inconsistent — especially with accents outside of São Paulo and Rio
- Google Recorder: excellent in PT-BR, but limited to a few phone models. And it doesn’t generate a summary, just linear transcription
- Neither one delivers speaker identification, automatic summaries, or topic-based organization. You get a continuous block of text
Sintesy fills this gap: it takes the same audio that the phone’s recorder produces, but returns something usable — not just text, but structure.
What About Consent? Can I Record Without Telling Anyone?
The short answer is: not recommended.
The LGPD (Brazil’s General Data Protection Law) classifies voice as personal data when it allows a person to be identified. Recording a meeting without consent may constitute irregular data processing, especially if the recording is used for purposes beyond internal record-keeping.
In practice, what works:
- Announce at the start of the meeting that you’ll be recording for transcription and minutes
- Make the purpose clear — “I’ll record this to generate the minutes, and then keep the audio for X amount of time”
- If someone objects, don’t record. Or record only your own speech as a personal note
- Set a retention period — don’t keep the audio indefinitely. Once it’s transcribed and the minutes are generated, the audio can be deleted
This isn’t bureaucracy. It’s the kind of care that prevents problems if the transcription is ever used as a reference in a discussion about what was decided.
When It’s Worth Using an External Microphone
For most meetings in a small room (up to 6 people), the phone in the center of the table does the job. But there are situations where an external microphone makes a difference:
- Large room (more than 8 people): the phone can’t pick up distant speakers well
- Meeting with constant background noise: open environment, café, coworking space
- Recurring meetings that you transcribe every week: the investment pays for itself quickly
It doesn’t need to be expensive. A Bluetooth lapel microphone or a USB tabletop microphone already significantly improves audio capture. The audio arrives cleaner to the AI and the transcription comes out with fewer errors.
Checklist: From Recorder to Ready Minutes
- Phone in the center of the table, microphone facing the participants
- Do Not Disturb mode on, Bluetooth off
- Notify participants about the recording
- Start recording on the native recorder (iPhone or Android)
- When finished, export the audio file
- Send the audio to Sintesy (app or dashboard)
- Receive transcription + automatic summary
- Review proper names and sensitive sections
- Adjust speaker attribution if necessary
- Save the final minutes and share with the team
- Set a deadline to delete the original audio
Conclusion
Transcribing in-person meetings was never difficult because of a lack of technology — it was difficult because of a lack of a clear path. The technology was scattered across recorders, foreign transcription services, and text editors. Bringing it all together into a workflow that works in Brazilian Portuguese was what was missing.
With the phone in the center of the table and the audio in Sintesy, the path becomes straightforward: recording, upload, transcription, summary, minutes. What used to be an hour of listening and typing becomes five minutes of review.
If you still rely on manual notes in in-person meetings, try it with your next one. Put the phone on the table, hit record, and see what changes when the text arrives ready.


