How to upload audio to ChatGPT for transcripts (2026): meeting notes, supported formats, size limits and which plan you need

What ChatGPT audio uploads do

OpenAI’s release notes for October 6, 2026 add “Audio uploads in ChatGPT”: attach an audio file to a conversation and ChatGPT can transcribe it, summarize it and answer questions about it. OpenAI’s own examples are turning a meeting, interview or lecture into structured notes, or drafting a follow-up email from it.

The Help Center article “Uploading files and audio to ChatGPT” describes four kinds of audio work:

• Transcribe: ask for a transcript, then ask follow-up questions about it.
• Summarize: summarize a meeting recording, highlighting decisions, open questions and next steps.
• Find things: ask “What did they say about the launch deadline?”, or pull requirements, examples and takeaways out of an interview or lecture.
• Transform: turn a recording into structured notes, a meeting recap or a follow-up email draft.

This is not voice mode. Voice mode is a live spoken conversation; audio uploads are for recordings you already have.

Who can use it: Free, Go, Plus and Pro

OpenAI’s wording is “available with paid ChatGPT subscriptions and workspaces, including Enterprise”, and “not available on the Free plan at this time”. So you need to pay, but OpenAI does not list the plans by name. Even on a paid plan, availability can vary by workspace settings, region, client version and selected model; if you do not see the option, update the app or try the web version.

About Go: it is a paid plan, but neither the Help Center nor the pricing page lists it for audio uploads, and the pricing comparison has no audio-upload row. The one recording-related row is ChatGPT record mode: No on Free and Go, Yes on Plus and Pro. That is a separate feature, so it shows Go and Plus differ around recording, not that Go lacks audio uploads. If you want to pay once and be sure, choose Plus.

Supported formats, file size and recording length

The upload-rate figures in the same Help Center article (80 files every 3 hours, 3 a day on Free) sit in the section on document and image limits, which says it covers documents, spreadsheets, presentations, text files and images. OpenAI publishes no separate upload count for audio, so do not assume those numbers apply.

Five steps from recording to transcript and notes

1. Check that you are on a paid plan: Open Settings → My Plan on chatgpt.com. OpenAI says audio uploads are available with paid subscriptions and workspaces and not on the Free plan at this time; if you are on Free, get Plus first.
2. Prepare the audio file: Use one of WAV, MP3/MPEG, OGG/OGA, audio-only WebM, PCM, FLAC, AAC, M4A and audio-only MP4, up to 512 MB per file. If you have a video (WebM and MP4), export an audio-only copy such as M4A or MP3 first: files identified as video are not accepted as audio uploads.
3. Attach it to a conversation: Start a new chat and pick the file with the upload (attach) control next to the message box. If the recording is sensitive, read the privacy section below before you upload.
4. Say exactly what you want: In the same message, ask for a transcript, summary, minutes or an email draft, and give speaker names, special terms and the output format. Seven ready-made prompts are below.
5. Review and follow up: OpenAI warns that transcripts may contain errors and speaker identification may be unreliable: check names, numbers, dates and decisions against the recording, then ask follow-up questions or request changes in the same chat.

Seven prompts you can copy

Swap in your own names and topics. The clearer you are about who is speaking, their roles and the special terms, the more usable the result.

This is an internal meeting recording. First give me a full transcript, then write minutes in this format:
1. Topic and attendees (Anna – product, Ben – engineering, Carla – sales)
2. Discussion points, grouped by agenda item
3. Decisions made
4. Action items as a table: task | owner | deadline (write "not mentioned" if nobody said)
5. Open questions
Mark any word you are unsure about as [?] instead of guessing.

This is a two-person interview. The interviewer is Sam; the guest is Dr. Lee. Transcribe it as a dialogue with the speaker's name before each turn.
Keep the wording, but drop filler words like "um" and "you know". Mark anything inaudible as [inaudible].
After the transcript, list the five lines that would work best as direct quotes.

This is a 50-minute lecture on inflation in macroeconomics. Turn it into revision notes:
- One-sentence summary of the lecture
- Key concepts in the order they were taught, each with a definition, an example and anything the lecturer stressed
- A separate list of formulas and technical terms
- Five self-test questions with answers

Listen to this recording and write two summaries of up to 150 words each, one in English and one in Spanish, with the same content.
Then add a two-column glossary of the technical terms used in the recording.
Keep people's and company names in their original spelling in both versions.

Based on this client meeting recording, draft a follow-up email to the client's project lead.
Tone: professional and brief. Thank them for the meeting, restate what we agreed in 3–5 bullet points, list our next steps and theirs, and propose a time for the next call.
Use only what was said in the recording; do not add commitments nobody made.

This is the full audio of a podcast episode. Write show notes:
- Three title options
- An episode description under 80 words
- 6–8 main topics in the order they come up
- Books, tools and websites the guests mention
- Three quotable lines for social media (exact words only)

No transcript yet. What does the recording say about the new release date? Who said it, and was a specific date given?
If people said different things at different points, list each version.

For prompt 4, replace Spanish with whatever second language you need.

Six habits for more accurate transcripts

1. Record it cleanly. Keep the phone close to the speakers and away from fans and keyboards; for online calls, use the meeting app’s own recording rather than re-recording a speaker.
2. Say who is talking. OpenAI says speaker identification may be unreliable. “Three people: A chairs, B is engineering, C is the client” beats letting ChatGPT guess.
3. Give it a glossary. List company names, product names, people and acronyms up front; misspelled names drop sharply.
4. Make it flag doubts. Ask it to mark unclear words as [?] rather than guess, then check only those spots.
5. Do not count on timestamps. Neither the Help Center nor the release notes mention them. Ask for rough positions (start, middle, end) if you need to find a passage, then listen to confirm.
6. Test a short clip, split long ones. Very long recordings may time out. Try 3–5 minutes first, check language and format, then upload the rest in parts.

Privacy: is your recording used for training?

• Turn training off: Settings → Data controls → Improve the model for everyone. With it off, new conversations are not used to train OpenAI’s models; your history stays.
• Temporary chat: stays out of history and memory and is not used to improve models, though it may be retained for up to 30 days for safety.
• Feedback counts: a thumbs-up or thumbs-down can bring the whole conversation into training even after you opt out.
• Workspaces: Business, Enterprise and Edu content is not used for training by default and follows the organization’s own retention settings.
• Get consent first: meetings, interviews and calls contain other people’s voices and information. Ask before you upload, follow your company’s policy and local law, and keep confidential recordings out.

Which plan is enough: Plus for most people, Pro 100 for heavy use

Plus is enough if you turn a few meetings, interviews or lectures a week into minutes, summaries and email drafts. Audio uploads are a paid feature and OpenAI publishes no audio difference between Plus and Pro, so for audio alone Plus is the lowest plan you can be sure of.

Consider Pro 100 if you work through recordings all day and run long analyses on the transcripts, so you keep hitting Plus model limits, or you want dozens of transcripts in one project (Plus allows 25 files per project, Pro 40). OpenAI does not say Pro gets more audio; what you pay for is higher model allowances and the Pro features. The full comparison is in ChatGPT Plus vs Pro 100.

Who does not need to pay for this

• You only need a raw transcript now and then. If you do not need summaries or follow-up questions, there are free ways to transcribe audio, from built-in phone and computer features to free tools. Start there.
• You need a verbatim, evidential record. For legal, medical or audit work, an AI transcript is a first draft; have a person check every line or use a professional transcriber.
• The recording must not leave your company. Follow the policy; do not send confidential audio to any outside service for convenience.
• You already have Plus or Pro. You can use it now; if the option is missing, update the app or try the web.

Getting Plus here: top-up or ready-made

In one line: keep your history with a top-up; take a ready-made account if you would rather not submit sign-in data or want a separate account for work recordings. The ready-made premium, about 2.09 USDT, applies to the first month only. The full buying guide is How to buy ChatGPT Plus.

Ordering the Plus top-up:

1. Check Settings → My Plan shows Free (or an expired Plus). If auto-renewal is on at the App Store or Google Play, switch it off there so you are not charged twice.
2. Choose ChatGPT Plus as a top-up on the pricing page. Following the order page, open chatgpt.com/api/auth/session in the same browser while signed in, copy all of the JSON and paste it into the order form, never into a chat.
3. Pay the exact USDT amount on TRC20 or BEP20 after the check passes; if it fails, you pay nothing. Delivery usually takes 5–30 minutes; confirm Plus under Settings → My Plan and try a short recording in a new chat.

Ready? ChatGPT Plus top-up, 27.58 USDT. New to USDT? The USDT payment guide covers network, exact amount and confirmation.

Is it legit? What is in writing, and what we cannot promise

• Checked before you pay. A top-up checks your sign-in data first and you pay nothing if it fails; the data is cleared after delivery and you keep your password.
• Refund terms are public. If an OpenAI regional policy change makes the service unavailable, you get a refund for the unused days, processed within 1 business day and returned in USDT; bans caused by breaking OpenAI’s usage policy are not refunded. Contact the bottom-right live chat promptly if anything goes wrong.
• Track record. ClaudeMax has operated since November 2024 and has served more than 20,000 users (our own figure).
• We are not OpenAI. ClaudeMax is independent, not OpenAI and not an authorized reseller. OpenAI decides where audio uploads are available and how much you can use; its terms prohibit sharing credentials and reselling access, so keep one person per account and change a ready-made account’s password at once. Renewals for ready-made accounts go through Telegram @ccnb1937.