How to Transcribe Audio with ChatGPT Using File Uploads
You can now transcribe audio with ChatGPT by uploading a recording and asking questions about it. A practical guide for journalists and students on paid plans.
I have a folder of voice memos on my phone that I never listened to again. I think many people have the same folder. Since October 6, you can transcribe audio with ChatGPT by attaching the file to a chat, and then ask questions about what was said.
What is new
OpenAI added audio uploads to ChatGPT. You attach a recording, ask for a transcript or a summary, and then keep asking follow-up questions. According to the official help page, files can be up to 512 MB. It accepts common formats like MP3, M4A, WAV, FLAC, OGG and audio-only MP4 or WebM. Video files are not supported, so export the audio first.
Two things to know before you try. First, it needs a paid plan (Plus, Pro or a workspace). It is not on the Free plan. Second, OpenAI says availability "may vary by workspace settings, region, client version, and selected model." I could not find a clear statement about Germany or the EU, so check in your own app.
Who this helps most
A freelance journalist. You record a 40 minute interview, and the hardest part is finding that one good quote again. Now you can ask "where did she talk about rent prices?" instead of scrolling through the audio.
A university student. Many lectures are recorded. You can turn a recording into clean notes, then ask the questions you were too shy to ask in the room.
How people are already using it
It is very new, so real stories are still few. A writer at Craving Tech said the first thing they want to do is upload old Zoom calls that were never transcribed, and ask things like where pricing was discussed. One early test on DEV Community, done on the Free plan, showed ChatGPT refusing to make a transcript instead of making one up. That fits the paid plan rule, and I like that it did not invent text.
How to transcribe audio with ChatGPT, step by step
- Get the recording as an audio file. From a video call, export the audio only.
- Open a new chat and attach the file with the paperclip.
- Write a clear prompt. For example:
Transcribe this interview. Label the speakers as Interviewer and Guest. Then give me the five strongest quotes and a short summary in five bullet points.
- Ask follow-ups, like "Which parts were hard to understand?" or "Write a short email to the guest with the two facts I still need to check."
For students, try this one:
Turn this lecture into study notes with clear headings. Then write 10 quiz questions about it, with answers at the end.
The honest limits
OpenAI itself says transcripts can have errors and speaker labels may be wrong. Quality also changes between languages, so a German recording may come out weaker than an English one. Very long files can time out. For anything you publish or quote, listen to that part of the original again. And only record people when they agree to it. In Germany this is not a small detail.
If you like practical AI tips like this, without the hype, subscribe to my newsletter. One entry a week, straight to your inbox.
Sources
If this was useful
Telc German B1 Exam Success Kit (20/80 Rule)
The 20% of German that keeps showing up in the telc B1 exam. Learn it first and walk in already knowing what to expect. €25