Gemini Meeting Minutes: Prompts and What the Free Tier Really Allows

The call ends. You export the recording. You open Google AI Studio. You paste the prompt you keep saved in a notes app. You copy the output and drop it into Slack.

Third time today?

This article covers the AI Studio workflow, prompts you can copy, what the official limits actually are, and how far the manual steps can be reduced.

There are two routes to Gemini meeting minutes

Sort this out first, or you will not know whether the advice you are reading applies to you. The two methods have completely different prerequisites.

Route 1: upload a recording to Google AI Studio

Anyone with a Google account can use it, and it does not care which meeting tool you used. The catch is that exporting the recording, uploading it, and entering prompts are all on you, every time.

Route 2: use the built in Gemini notes in Google Meet

Google Meet has Gemini note taking built in. The meeting ends and the notes already exist.

Which one applies to you

Everything that follows covers route 1.

The workflow: meeting minutes in Google AI Studio

What you need

• A recording: supported formats include MP3, WAV, AIFF, AAC, OGG, FLAC, and M4A
• A Google account

When exporting audio from a meeting recording, choose MP3. The reason will become clear in the limits section.

The steps

1. Have your recording ready
2. Open Google AI Studio
3. Upload the file
4. Ask for a transcript
5. Then ask it to turn that text into minutes

Why you should not ask for minutes in one shot

Two reasons. First, doing both at once lets summarization take over, and specifics from the discussion get dropped. Second, keeping the intermediate transcript means you can go back later and check whether a decision was actually stated.

For minutes that carry commitments to a customer, being able to verify matters more than it sounds.

Prompts you can copy

Step 1: transcription

Transcribe this audio as faithfully to what was said as possible.
Where speakers can be inferred, label them as "Speaker A", "Speaker B", and so on.
Do not summarize or rephrase. Write out what was actually said.

Step 2: turning it into minutes

Using the transcript above, produce meeting minutes.
Use these sections:

- Overview
- Decisions made
- Open items
- Action items (state owner, deadline, and definition of done)
- Key discussion points
- To confirm next time

Do not include anything the transcript does not support.
Where you cannot determine something, write "needs confirmation".

Those last two lines matter more than they look. Without them, the model fills gaps from context. Minutes are exactly the kind of document where filled gaps cause problems, so rule it out explicitly.

Adjusting by meeting type

The limits: AI Studio handles a one hour meeting fine

The numbers floating around online contradict each other, so here is the official specification.

What Google AI Studio allows

AI Studio runs on the Gemini API specification. The official documentation states:

About that "five minute free limit"

What still deserves attention

Generous limits do not mean nothing to watch for.

• Export audio as MP3. Uncompressed WAV runs to several hundred megabytes per hour and takes a while to upload. The same meeting as MP3 lands in the tens of megabytes
• Longer audio costs more tokens. An hour of meeting audio is roughly 115,000 tokens. Using the free tier, you may hit a rate limit
• Splitting is generally unnecessary. Split the file and a concern raised in the first half, resolved in the second, becomes invisible to the model. With this much headroom, hand it over in one piece

The sharing step stops you every time

You have minutes now. The work is not finished.

Producing them is not delivering them

Copy the output, paste it somewhere, share that URL. That last chore recurs every time.

Count the steps

At six meetings a week, here is what one meeting costs:

1. Export the recording
2. Upload to AI Studio
3. Paste the transcription prompt
4. Paste the minutes prompt
5. Copy the output
6. Paste it into wherever you share

Six steps. At six meetings a week that is 36 a week, roughly 150 a month. At a minute each, two and a half hours a month disappears.

How far the manual steps can be cut

A few meetings a month? Stay manual

There is no need to add another tool.

Past a certain frequency, better prompts stop helping

The problem arrives with volume.

Same Gemini, different route

One fact worth knowing here, because it changes what your options are.

If you are unsure about the data

Some people are reluctant to send recordings to an external AI, and reasonably so. Some AI notetakers may use submitted data for training, which is a real thing to check.

Summary: get the workflow right, then decide on frequency

• Two routes exist. On Business Standard or above, the built in Google Meet Gemini notes are faster. Otherwise, upload to AI Studio
• Split transcription and summarization into two steps. More reliable, and verifiable afterward
• Always include "do not include anything the transcript does not support" and "write needs confirmation where undetermined"
• AI Studio takes up to 9.5 hours of audio per prompt, so a 60 minute meeting is fine. The "five minute free limit" is the Gemini app, not AI Studio
• Export audio as MP3, and hand the meeting over in one piece rather than splitting it
• A few times a month, manual is fine. At several meetings a week, decide by counting steps

Start by locking in the workflow with the prompts above. Then count your monthly steps and see whether the math still works.

Qureco Screen Recorder

Powerful screen recording app for Mac

Record meetings, let AI handle the notes, just read what arrives in Notion.
Try all features free for the first month.

No Setup RequiredNo WatermarkAI Meeting NotesNotion Integration

Download for FreeLearn More

Frequently asked questions

Using Google AI Studio, yes, as far as cost goes. Audio runs up to 9.5 hours per prompt, so a normal meeting will not approach the ceiling. You may hit a rate limit on the free tier, in which case waiting a short while lets you continue. The "five minute" figure applies to the Gemini app and not to AI Studio.

Not automatically. Even with prompting you get labels like "Speaker A" and "Speaker B". Real names require a find and replace pass afterward. If people introduce themselves at the start of the recording, adding that context to your prompt improves how reliably the labels map to people.

The AI Studio route does. Any recording works regardless of the meeting tool. The built in Gemini notes are Google Meet only, so if your calls are mostly Zoom or Teams, route 1 is your option.

Audio quality, ahead of any prompt change. A recording where each participant speaks into their own microphone transcribes noticeably better than one picked up through a conference room speaker. Splitting transcription from summarization is the next biggest factor.

Glossary