Can ChatGPT Transcribe Audio & Video? (Honest 2026 Answer)

Quick answer

Partially. ChatGPT Plus / Pro / Business transcribe audio two ways: Record mode (macOS desktop only, 4-hour session cap) captures live meetings and voice notes; GPT-4o file upload accepts MP3/WAV/M4A up to 25 MB (~10-15 min). ChatGPT Free cannot transcribe audio. No ChatGPT tier plays video files — extract audio first (FFmpeg, yt-dlp). Developers use the OpenAI API at $0.006/min (whisper-1) or $0.003/min (gpt-4o-mini-transcribe).

The 3 methods, ranked by usefulness

"ChatGPT transcription" means three different things depending on which product surface you're using. Ranked by how often each one is the right call:

1. Record mode

Availability: Plus / Pro / Business / Enterprise / Edu — macOS desktop app only

Use case: Live meetings, brainstorms, voice notes captured in the moment

Limits: 4-hour session cap (auto-stop), no speaker labels, English at launch

Verdict: Works well for solo capture + single-speaker interviews on macOS

2. File upload (GPT-4o)

Availability: Plus / Pro / Business — all platforms

Use case: Transcribing existing audio files (MP3, WAV, M4A, WebM)

Limits: 25 MB per file (~10-15 min at typical bitrates), ~80-86% accuracy, no timestamps

Verdict: Works in practice but not officially documented by OpenAI — behavior may change

3. Voice dictation

Availability: All tiers — mobile + desktop microphone button

Use case: Dictating a message into a chat

Limits: Not for transcribing existing files — mic input only

Verdict: Common misunderstanding — this is speech-to-input, not audio-to-transcript

Capability by ChatGPT tier

What each ChatGPT plan can and cannot do for transcription. Verified against OpenAI's ChatGPT Record help article, chatgpt.com/pricing, and MacRumors' 2025-07-17 Plus rollout confirmation.

Tier	Record mode	Audio file upload	Video direct
ChatGPT Free	❌ Not available	Text-only (PDF, DOCX, images, XLSX, TSV, CSV)	❌ No
ChatGPT Plus ($20/mo)	✓ macOS desktop only	✓ MP3/WAV/M4A/WebM up to 25 MB (undocumented but works)	❌ No — extract audio first
ChatGPT Pro ($200/mo)	✓ macOS desktop only	✓ Same as Plus + higher usage caps	❌ No — extract audio first
ChatGPT Business ($25/seat)	✓ macOS desktop only	✓ Same as Plus + workspace controls	❌ No — extract audio first
ChatGPT Enterprise / Edu	✓ macOS desktop — admin-configurable per workspace	✓ Admin-configurable	❌ No — extract audio first
OpenAI API (developer)	N/A — direct model access	✓ 25 MB per request	MP4 accepted but audio-track only

Common misunderstanding: "ChatGPT can transcribe" conflates the Record feature (macOS desktop live capture) with the GPT-4o file upload (existing audio files, undocumented but functional) with the OpenAI API (developer endpoint, separate product). Different limits, different platforms, different results.

The Record feature — deep dive

ChatGPT Record is OpenAI's built-in live audio capture, rolled out to Plus users globally on July 17, 2025. It captures meetings, brainstorms, lectures, and voice notes in real time and produces a transcript plus AI-generated notes.

Availability (verified 2026-08-15)

macOS desktop app only. Windows support has no announced roadmap. Web ChatGPT does not have Record.
Plans: Plus, Pro, Business, Enterprise, Edu. Not available on Free.
Enterprise admins can disable Record per workspace.

Session limits

4-hour cap (240 min) per session. Sessions exceeding this stop automatically and generate notes uploaded as a private canvas.
No hard file-size limit for the resulting transcript (it's stored in the chat).
Language support: English at launch, expanding gradually.

Output

Transcript — single-track prose, no speaker labels.
Summary — auto-generated overview.
Time-stamped citations — clickable references from the summary back into the transcript. This is the differentiator vs plain transcription tools.
Suggested follow-ups — action items and next-step prompts.
All output appears in a canvas you can edit or export.

Privacy

Audio recordings are deleted after transcription. Only the transcript persists (in your chat history until you delete it).
OpenAI does not train models on Record audio. Confirmed in the help article.
Team/Business/Enterprise: recordings not used for training by policy.
Plus/Pro: same non-training default, plus standard opt-out controls apply.

Steps

Open the ChatGPT macOS desktop app.
Start a new chat. Click the Record icon in the input bar.
Grant microphone permission. Recording starts. Audio is captured on-device and streamed to OpenAI for real-time transcription.
Speak, or let the meeting audio play through your Mac speakers (Record captures system audio too).
Click stop. The canvas opens with transcript + summary + citations + follow-ups.
Ask ChatGPT to reformat, extract action items, translate, or restructure — all in the same chat.

Transcribing existing audio files (GPT-4o upload)

Since GPT-4o launched in 2024, ChatGPT Plus / Pro / Business accept audio file uploads directly in chat. It processes the audio through Whisper server-side and returns a transcript in the chat. Important caveat: as of 2026-08-15, OpenAI's official "supported file types" documentation covers only text formats (PDF, DOCX, XLSX, CSV, TSV, PPTX, TXT). Audio upload works in practice but is not formally documented — meaning behavior could change without notice.

What works

Formats: MP3, WAV, M4A, WebM. MP3/M4A most consistent.
Size limit: 25 MB per file (matches the OpenAI API limit).
Duration: ~10-15 minutes of typical-bitrate audio fits under 25 MB.
Output: prose transcript in chat, no timestamps, no speaker labels.
Post-processing: same chat can then summarize, translate, extract action items.

What doesn't work well

Files over 25 MB — hard fail.
Files under 25 MB but over ~10 minutes — often returns incomplete transcripts without warning.
Multi-speaker audio — no diarization.
Accented English, overlapping speech, technical vocabulary — accuracy drops below 80%.
Non-English audio — supported by Whisper but Record UI + processing prioritizes English.

How to split larger files

For files over 25 MB or longer than 10 minutes, split with FFmpeg into ~10-min chunks:

ffmpeg -i input.mp3 -f segment -segment_time 600 -c copy chunk%03d.mp3

Upload each chunk separately to ChatGPT, then concatenate the transcripts. Or use the OpenAI Whisper API which handles the same 25 MB limit per request but is designed for chunking workflows. For hands-off single-upload processing of long files, a hosted transcription service handles this natively.

Transcribing video with ChatGPT

No ChatGPT tier plays video files directly. Not Free, not Plus, not Pro. Uploading an MP4 to a chat either fails or extracts nothing useful. The OpenAI API accepts MP4 but transcribes only the audio track — the visual content is ignored. Every "ChatGPT transcribes video" workflow is really a two-step: extract audio, then transcribe audio.

Standard workflow

Extract audio from the video:
- FFmpeg (free CLI): ffmpeg -i video.mp4 -vn -c:a copy audio.m4a
- yt-dlp for YouTube: yt-dlp -x --audio-format mp3 [URL]
- QuickTime on Mac: File → Export As → Audio Only
- Online: audio-extractor.net, Kapwing extract-audio
Verify the audio file is under 25 MB and ~10 min.
Upload to ChatGPT Plus/Pro/Business, or use Record playback capture, or send to the OpenAI API.
Ask ChatGPT to reformat, summarize, extract action items.

Shortcut — skip the two-step

Dedicated transcription services read the video directly: VexaScribe accepts MP4/MOV/MKV/WebM/AVI up to 5 GB, extracts audio internally, transcribes with Whisper large-v3, and returns the transcript with speaker labels — one upload, one download, no CLI. For iPhone videos specifically see transcribe iPhone video; for adding captions afterward see add SRT to MP4.

The realistic workflow — post-hoc, not live

Every ChatGPT-transcription review that actually tested the product (tldv, meetgeek, jotme, umevo) lands on the same conclusion: the realistic workflow for a meeting is after the fact. Record the call somewhere else, export the file, upload it, and hope it comes back usable.

Record mode helps in a narrow band: solo brainstorms captured on macOS, single-speaker interviews under 4 hours, voice notes. It falls apart for the workflows people actually want to use it for.

Where Record actually shines

Solo brainstorm or voice memo on a Mac (single speaker, English, under 4h)
1-on-1 interview where you don't need speaker labels (mark them by hand later)
Short lectures or study capture where AI-generated notes are the primary output
Ideation sessions where the summary matters more than the raw transcript

Where Record falls apart

Multi-speaker meetings — no diarization means everyone becomes one voice
Long recordings over 4h — session auto-stops mid-flow
Non-English audio — support is expanding but launch was English-only
Windows / mobile — no support as of 2026-08-15
Batch processing — no CLI, no API, no queue
Zoom/Teams/Meet calls — Record captures system audio locally, but purpose-built meeting bots (Otter, Fireflies, Fathom) integrate directly with the calendar and produce speaker-labeled notes

Prompts that work with ChatGPT + a transcript

ChatGPT's real strength is post-processing existing transcripts — cleanup, formatting, summarization, translation. It is not the transcription engine; Whisper (or whatever ASR you used) is. Five prompt patterns that reliably produce useful output:

Add punctuation and paragraphs

You will receive an unpunctuated transcript. Add standard sentence punctuation, capitalize proper nouns and sentence starts, and break the text into paragraphs where the speaker changes topic. Do NOT change any words. Do NOT add facts. Return only the cleaned text.

Why it works: ChatGPT excels at post-processing existing text. This prompt keeps hallucination risk low by forbidding word changes.

Infer speaker labels (2-person interview)

This is a two-person interview transcript. Label each turn with Speaker 1: or Speaker 2:. If you cannot tell who is speaking, mark the turn [SPEAKER ?]. Do not invent dialogue.

Why it works: GPT-4 can infer speaker changes from linguistic cues (question-answer patterns, name references). The [SPEAKER ?] escape hatch prevents fabrication. Fragile beyond 2 speakers.

Enforce a glossary

Below is a transcript that references these terms: [Vercel, Next.js, Turbopack, App Router]. Correct any misspellings of these terms to match this exact casing. Leave all other text unchanged.

Why it works: Whisper often mangles technical vocabulary. ChatGPT can fix domain-specific terms without touching the rest.

Flag low-confidence sections

In this transcript, find any sentence that looks garbled, contains repeated tokens, or has numbers/proper nouns that seem out of context. Wrap each in [FLAG: reason]. Return the full transcript with flags inline.

Why it works: Turns ChatGPT into a QA pass. Human reviewer then only reads the flagged sections.

Extract action items with quoted context

From this meeting transcript, list every action item as: '[Owner if named]: [action], because [1-sentence quote from transcript]'. If owner is unclear, mark [UNASSIGNED]. Do not invent action items — only extract what is explicitly discussed.

Why it works: Meeting notes are the #1 real ChatGPT use case. Requiring quoted context forces grounding.

Prompts that fail (and what to do instead)

Common requests that don't work — and the actual working path for each.

"Transcribe this YouTube video" (paste URL)

What people try: Sending a URL and asking ChatGPT to transcribe.

What actually happens: ChatGPT cannot fetch video URLs or play video files. Even with browsing enabled, it can read the page metadata — not the audio track. On Free tier, it declines or hallucinates a plausible-sounding fake transcript. On Plus, it politely refuses.

Fix: Extract audio first (yt-dlp: yt-dlp -x --audio-format mp3 [URL]), then upload the MP3 (Plus/Pro) or use the OpenAI API.

"Upload this 45-minute podcast" (large file)

What people try: Uploading a large MP3 (over 25 MB or over ~10 min) to a Plus chat.

What actually happens: Upload fails past 25 MB. Files under 25 MB but longer than ~10 min often return incomplete transcripts — ChatGPT stops mid-file without warning.

Fix: Split the file with FFmpeg (ffmpeg -i input.mp3 -f segment -segment_time 600 -c copy chunk%03d.mp3) and process chunks separately. Or use the OpenAI API at $0.006/min. Or use a hosted transcription service that handles long files natively.

"Transcribe verbatim, including every uh and um"

What people try: Asking ChatGPT to preserve filler words from an already-cleaned transcript.

What actually happens: Whisper drops most filler words during transcription. ChatGPT cannot add them back — the information is gone.

Fix: Configure Whisper at the transcription step with the correct decode parameters, not after. faster-whisper's word_timestamps=True + no VAD filtering preserves more fillers.

"Fix the timestamps in this SRT"

What people try: Asking ChatGPT to re-time SRT cues to match audio.

What actually happens: ChatGPT cannot listen to audio and cannot align text to waveform. It will invent plausible-looking timestamps that drift by the end of the file.

Fix: Re-transcribe with WhisperX or faster-whisper (both produce word-level timestamps), or manually adjust in Subtitle Edit / Aegisub.

"Diarize this transcript — say who is speaking"

What people try: Asking Record output or a Whisper transcript to be split by speaker.

What actually happens: For 3+ speakers with no linguistic cues (no name references, no distinct speech patterns), ChatGPT guesses. Guesses look confident and are wrong.

Fix: Use a diarization-enabled service at the transcription stage (VexaScribe, Otter, Rev, AssemblyAI, Deepgram). Diarization is an audio-processing task, not a text task.

File-size, duration limits + OpenAI API pricing

Limits by interface

Interface	Max file size	Max duration	Notes
Record (Plus/Pro/Business/Enterprise)	N/A (live capture)	4h per session (auto-stop)	macOS desktop only
GPT-4o audio upload (Plus/Pro/Business)	25 MB	~10-15 min effective	MP3/WAV/M4A/WebM, undocumented feature
Text file upload (all tiers)	512 MB	N/A	PDF/DOCX/XLSX/etc. — no ASR
OpenAI API /v1/audio/transcriptions	25 MB per request	~25 min at 128 kbps MP3	Split longer files client-side

OpenAI API pricing (transcription models, verified 2026-08-15)

Model	Per minute	Per hour	Notes
whisper-1	$0.006	$0.36	Original Whisper large-v2 hosted by OpenAI. Best for long-form (podcasts, lectures, calls).
gpt-4o-transcribe	$0.006	$0.36	GPT-4o audio branch. Best for short conversational turns. Weaker than whisper-1 on long-form per benchmarks.
gpt-4o-mini-transcribe	$0.003	$0.18	Cheapest tier. Best for high-volume simple audio.
gpt-4o-realtime (WebSocket)	$0.017	$1.02	Live streaming transcription. Used for real-time voice apps.

Per-minute billing, no minimum, no publicly-listed volume discount. For long-form audio (podcasts, meetings, earnings calls), whisper-1 significantly outperforms gpt-4o-transcribe per Novascribe's 904-file benchmark. See how accurate is Whisper for the full comparison data.

Accuracy — honest limits

Every ChatGPT-transcription review that measured accuracy lands around 80-86% on clean audio, dropping significantly for real-world audio (accents, overlapping speakers, technical vocabulary, background noise). That's below Whisper large-v3 self-hosted (~95%) and below every purpose-built transcription API.

Engine	Clean audio	Real-world	Diarization	Note
ChatGPT Record / GPT-4o upload	80-86%	70-80%	❌ None	Whisper large-v2 under the hood, plus GPT-4o post-processing
Self-hosted Whisper large-v3	95-97%	88-92%	❌ (add pyannote separately)	Reference model — best raw accuracy for open-source
Whisper API (whisper-1)	94-96%	85-91%	❌ (API-side)	Hosted Whisper large-v2. Long-form beats gpt-4o-transcribe significantly
VexaScribe (Whisper large-v3 + diarization)	95-97%	88-92%	✓ Up to 50 speakers	Hosted, includes speaker labels, SRT/VTT/TXT export
Deepgram Nova-3	97%	90-93%	✓ Included	Commercial API, low latency
AssemblyAI Universal-2	97%	90-93%	✓ Included	Commercial API, custom vocabulary

Accuracy numbers verified against independent testing by jotme.io, meetgeek.ai, and videotobe.com in 2026, plus Novascribe's own July 2026 benchmark of 904 audio files. ChatGPT's ~80-86% ceiling is fine for informal use; for anything shipping to a client, podcast RSS, or legal record, use a purpose-built engine.

When to skip ChatGPT for transcription

ChatGPT is a good post-processor and a fine live-recording tool for macOS Plus/Pro/Business users. It is not the right first choice when any of the following applies:

You have existing audio or video files over 25 MB. Use a dedicated service or the OpenAI API with chunking.
You need speaker labels. ChatGPT Record produces single-track output; dedicated services (VexaScribe, Otter, Rev, AssemblyAI, Deepgram) handle diarization.
Your audio is longer than ~2 hours. API caps requests at 25 MB; Record caps at 4h. Neither handles 8h all-hands or day-long court testimony gracefully.
You need SRT / VTT / word-level timestamps. ChatGPT output is prose; dedicated tools produce subtitle formats.
You're on ChatGPT Free. Record isn't available; you're restricted to pasting existing transcripts.
You're on Windows. Record has no announced Windows roadmap.
You're processing high volumes. Per-minute API pricing is cheap for individuals, expensive for batch operations; self-hosted Whisper or a hosted subscription is more predictable.
Meeting transcription with participants. Purpose-built meeting bots (Otter, Fireflies, Fathom, tldv) integrate with your calendar, join Zoom/Teams/Meet as a participant, and produce speaker-labeled notes.

VexaScribe covers audio and video transcription with speaker labels (up to 50 speakers), SRT/VTT/TXT export, and 100+ languages — no per-minute API billing, no chunking, no prompt engineering. Free tier available.

Verified sources

Every capability, limit, and price on this page was cross-checked against these sources on 2026-08-15:

OpenAI Help Center — ChatGPT Record (feature, tier availability, privacy)
chatgpt.com/pricing (Free / Plus / Pro / Business / Enterprise / Edu plans)
MacRumors — ChatGPT Plus Gets Record Mode on Mac (July 17, 2025 global Plus rollout)
AIbase — ChatGPT Voice Recording Mode Now Fully Available (macOS-only scoping)
OpenAI API pricing (whisper-1, gpt-4o-transcribe, gpt-4o-mini-transcribe)
Accuracy comparison synthesized from independent tests by jotme.io, meetgeek.ai, videotobe.com, and Novascribe's July 2026 904-file benchmark
Session cap (4h auto-stop) confirmed via Sally.io, tldv.io, and Noota independent reviews

Frequently asked questions

Can ChatGPT transcribe audio?

Yes, in three ways. (1) Record mode on ChatGPT Plus/Pro/Business — macOS desktop only, 4-hour session cap, captures live audio and produces a transcript with summary and time-stamped citations. (2) GPT-4o file upload on Plus/Pro/Business — accepts MP3/WAV/M4A/WebM up to 25 MB (~10-15 min). Not officially documented but works in practice. (3) OpenAI API — $0.006/min for whisper-1 or gpt-4o-transcribe, $0.003/min for gpt-4o-mini-transcribe. ChatGPT Free cannot transcribe audio files. Verified 2026-08-15.

Can ChatGPT transcribe video?

No ChatGPT tier plays video files directly. Uploading MP4 to a chat either fails or ignores the audio. The OpenAI API accepts MP4 but transcribes only the audio track — visual content is ignored. Working workflow: extract audio first (FFmpeg: ffmpeg -i video.mp4 -vn -c:a copy audio.m4a, or yt-dlp for YouTube), then upload the audio to ChatGPT Plus (under 25 MB) or use the API. For hands-off single-upload video-to-text, use a dedicated service like VexaScribe that reads MP4/MOV/MKV directly.

Can ChatGPT transcribe video to text?

Not directly — no ChatGPT interface plays video. Two-step workflow: (1) Extract audio with FFmpeg, yt-dlp, or QuickTime Export → Audio Only. (2) Send the audio to ChatGPT Plus/Pro/Business via Record playback capture (macOS), GPT-4o upload (all platforms, 25 MB limit), or the OpenAI Whisper API. For long videos or multi-speaker content, use a purpose-built service — ChatGPT lacks diarization and struggles with files over ~10 minutes.

How to ask ChatGPT to transcribe a video?

Skip the direct ask — ChatGPT will decline or hallucinate. Working path: (1) extract audio with FFmpeg (ffmpeg -i video.mp4 -vn -c:a copy audio.m4a), (2) if under 25 MB upload to ChatGPT Plus/Pro/Business directly, or use Record mode on macOS while playing the audio, or send to the OpenAI API at $0.006/min. (3) then paste-or-continue in the same chat to summarize, extract action items, or reformat.

Can ChatGPT convert videos to text?

Not directly — no ChatGPT interface plays video. Use a two-step: extract audio (FFmpeg, yt-dlp, QuickTime Export → Audio Only), then transcribe via ChatGPT Record (macOS Plus/Pro/Business), GPT-4o upload (Plus/Pro/Business, 25 MB limit), the OpenAI Whisper API ($0.006/min), or a dedicated service that handles both steps in one upload.

Is there an AI that can transcript a video?

Yes — dedicated transcription services (VexaScribe, Rev, Otter, Descript, AssemblyAI, Deepgram) accept video files directly, extract audio internally, and return transcripts with speaker labels and timestamps. Most run OpenAI Whisper large-v3 or a comparable model under the hood. Direct-to-video is what ChatGPT lacks; dedicated tools have it. For iPhone videos specifically see the iPhone video transcription guide.

Can ChatGPT read videos?

No — ChatGPT cannot decode or play video files (MP4, MOV, MKV, WebM). Uploading a video to a chat does nothing useful. You must extract the audio track first (FFmpeg -vn flag, yt-dlp -x, or online extractor), then transcribe the audio via ChatGPT Plus upload, Record mode, or the OpenAI API. ChatGPT can then process the resulting transcript text.

Can ChatGPT convert audio to text?

Yes on Plus/Pro/Business tiers. Two paths: Record mode (macOS desktop, 4-hour session cap, live capture) and GPT-4o file upload (all platforms, 25 MB / ~10-15 min per file, undocumented but functional). Free tier cannot. Developers use the OpenAI API at $0.006/min for whisper-1 or $0.003/min for gpt-4o-mini-transcribe. Accuracy ~80-86% on clean audio, dropping significantly for accented, multi-speaker, or technical audio.

Can ChatGPT transcribe audio for free?

No — ChatGPT Free does not include audio transcription. Free workarounds that produce a transcript you can then paste into Free-tier ChatGPT for cleanup: self-hosted Whisper (free, requires install), free tiers of hosted services (VexaScribe includes 30 minutes; others vary), or OpenAI API's ~$0.006/min minimum cost.

Which iPhone models support the built-in Notes transcript?

Not applicable to ChatGPT — that's iOS 18 Apple Notes, not ChatGPT. If you meant ChatGPT on iPhone: the ChatGPT iOS app supports voice dictation on all tiers (dictate into a chat) and displays Record output from macOS sessions synced via chat history. ChatGPT Record itself is macOS desktop only as of 2026-08-15; no iOS Record client exists.

Does the ChatGPT Record feature work on Windows?

No, not as of 2026-08-15. ChatGPT Record is macOS desktop only. OpenAI has referred to "future expansion" without an announced date for Windows or web-app support. Windows Plus/Pro/Business users can still use GPT-4o audio file upload (Plus/Pro/Business tiers, MP3/WAV/M4A/WebM up to 25 MB), the voice dictation microphone button in mobile/desktop chats, or the OpenAI API directly.

How long can ChatGPT Record sessions be?

4 hours (240 minutes) per session. Sessions exceeding this cap stop automatically and generate notes uploaded as a private canvas. No hard file-size limit on the resulting transcript. For 6-8+ hour capture (all-hands, day-long conferences), split the session manually or use a purpose-built meeting recording tool.

Are ChatGPT Record audio files stored by OpenAI?

No — per OpenAI's help documentation, Record mode audio recordings are used only for transcription and deleted afterward. Only the transcript persists in your chat history until you delete it. OpenAI does not train models on Record recordings. Team, Business, and Enterprise plans have contractual non-training defaults; Plus and Pro follow standard ChatGPT privacy settings with opt-out controls available.

What's the difference between ChatGPT Record and the OpenAI Whisper API?

Record is a live-capture consumer feature in the ChatGPT macOS desktop app (Plus/Pro/Business/Enterprise/Edu). The OpenAI Whisper API is a developer endpoint (api.openai.com/v1/audio/transcriptions) that accepts existing audio files up to 25 MB and returns text — no ChatGPT interface required. Record is for meetings you're in right now; the API is for batch processing existing files. The API supports whisper-1, gpt-4o-transcribe, and gpt-4o-mini-transcribe models at different price points.