How to Summarize a Recording With ChatGPT: Transcribe First, Then Prompt

You have a recording. Maybe it is a team meeting, a customer interview, a class, or a voice memo you made in the car. You want a short summary: what was decided, who does what, what to remember. ChatGPT is good at that part. But it works best when you give it text, not audio. This guide shows a simple two-step way: turn the recording into a transcript first, then hand the transcript to ChatGPT with a clear prompt.

Every limit in this article comes from a public page we checked on 28 October 2026. Links are next to each number.

Why text first, then ChatGPT?

You can upload some audio straight to ChatGPT. So why add a step? Three plain reasons.

• Plan limits. OpenAI's file uploads help page says audio uploads are available on paid ChatGPT plans and are not available on the Free plan. The same page says Free users can upload 3 files per day. A pasted transcript is not a file upload at all.
• You can check the words. The same OpenAI page warns that transcripts may contain errors and speaker identification may be unreliable. When you have your own transcript, you can read it, fix a name, and delete a private aside before anything goes into a chat.
• You can reuse it. One transcript can become a summary, an email, a study sheet, and a list of quotes. You do not need to process the audio again each time.

Step 1: Get a clean recording

Good summaries start with clear audio. A few quick habits help:

• Put the phone or laptop close to the person who talks most.
• Ask people not to talk over each other during decisions.
• Say the date and topic at the start. "Tuesday product sync, launch dates." It helps you later, and it helps the summary.
• Tell everyone you are recording. Rules about consent differ by place, so check yours.

Common sources and what they give you:

• iPhone Voice Memos: an .m4a file.
• Zoom, Meet, or Teams recording: usually an .mp4 video, sometimes a separate audio file.
• Lecture recorder or phone app: often .mp3, .wav, or .m4a.

Step 2: Turn the recording into text

Open the right page for your file. For a call, use meeting transcription. For a phone memo, use M4A to text. For a screen or Zoom recording, use video to text. For a class, lecture transcription works the same way. For a conversation with one guest, try interview transcription.

Here is what the free local mode gives you, from the whisperweb.dev product facts:

• Each file can be up to 20 minutes and 200 MB.
• No sign-up, and no daily file limit.
• The file is processed in your browser. In free local mode the audio does not leave your device.
• 100+ spoken languages.
• Export as TXT, JSON, SRT, or VTT.

Longer than 20 minutes? Split it at natural pauses and transcribe each part. Our guide on splitting a long recording shows how. If you do this every week with long files, the paid Unlimited plan uses the cloud and takes files up to 10 hours or 5 GB each. That is optional. The free path is enough for most short meetings and memos.

Export the transcript as TXT. Plain text is the easiest thing to paste into ChatGPT.

Step 3: Clean the transcript in two minutes

Do not skip this. It is the cheapest way to get a better summary.

1. Add speaker names. The transcript is one block of text. It does not label speakers for you. Add "Ana:" or "Teacher:" in front of key lines, at least for decisions.
2. Fix names and terms. Product names, people, and jargon are where speech models slip most. Search for them and fix them.
3. Cut what should not be shared. A salary number. A health detail. A side joke about a client. Delete it now, before it goes to any chat.
4. Add one line of context at the top. For example: "Weekly sync, 5 people, goal: agree on the October launch."

Step 4: Paste into ChatGPT with a clear prompt

Short transcripts can be pasted straight into the chat. Long ones can be saved as a .txt file and attached. OpenAI's help page lists TXT as a supported document type and caps text files at 2 million tokens per file. Most meeting transcripts are far below that.

Then ask for exactly what you need. Vague prompts get vague summaries. Here are prompts that work, by situation.

Meeting

"Summarize this meeting transcript. Give me: 1) decisions made, 2) action items with owner and due date, 3) open questions. Use bullet points. If an owner or date is not said, write 'not stated'. Do not guess."

Example: in a 15-minute sync, you might get three decisions, four tasks, and one open question about budget. That fits in a chat message to the team.

Interview

"This is an interview with a customer. List the five main points they made, each with one short direct quote from the transcript. Then list any problems they mentioned with our product."

Asking for quotes keeps the summary honest. You can search the transcript for each quote and check it.

Class or lecture

"Turn this lecture transcript into study notes. Include key terms with simple definitions, the main argument in three sentences, and five practice questions."

Example: a 20-minute biology class becomes a one-page sheet you can review before the quiz.

Voice memo

"These are my spoken notes, a bit messy. Turn them into a clean to-do list and a short paragraph of ideas. Keep my wording where you can."

Good for ideas you said out loud on a walk and never want to retype.

Step 5: Check the summary against the transcript

A summary can sound sure and still be wrong. Do a fast check:

• Pick each action item and search the transcript for it. Is it really there?
• Check every number and date. These matter most and are easy to mix up.
• Ask a follow-up: "Which parts of your summary are not directly supported by the transcript?"

If something looks off, go back to the transcript, not the audio. That is the point of having text.

Tips that save time

• Long meeting, many parts? Summarize each part first, then paste the part summaries together and ask for one final summary.
• Different language? Transcribe in the spoken language, then ask ChatGPT to summarize in the language your team reads.
• Same format every week? Save your favorite prompt in a note and reuse it, so every summary looks the same.
• Need subtitles too? Export SRT from the same transcript. One recording, two results.

What this workflow does not do

Be clear about the limits. Free transcription happens on your device, but once you paste text into ChatGPT, that text goes to OpenAI under your ChatGPT account settings. That is why Step 3 (cutting private lines) matters. The transcript also has no automatic speaker labels, and the free in-browser recorder uses your microphone, not the other side of a call. Record calls in the meeting app, then open the file.

Quick FAQ

Can I summarize a recording with free ChatGPT?

Yes, if you give it text. OpenAI's help page says audio uploads are not on the Free plan, but you can paste a transcript into a chat.

How long can my recording be?

On the free local mode of whisperweb.dev, each file can be up to 20 minutes and 200 MB, with no daily file limit. Split longer files, or use Unlimited for files up to 10 hours or 5 GB.

Do I need an account?

Not for free transcription on whisperweb.dev. ChatGPT needs its own account.

Is it the same as whisperweb.app?

No. The official site is whisperweb.dev. Sites named whisperweb.app or whisperweb.tech are not the same company.

Sources

• OpenAI Help Center, Uploading files and audio to ChatGPT (checked 28 October 2026): audio uploads on paid plans only, Free users 3 file uploads per day, text files capped at 2 million tokens, TXT supported, transcripts may contain errors.
• whisperweb.dev, product facts (llms.txt) (checked 28 October 2026): free local mode 20 minutes / 200 MB per file, no sign-up, no daily file limit, 100+ languages, TXT/JSON/SRT/VTT export; Unlimited cloud up to 10 hours or 5 GB per file.