Can you upload an MP4 or other video to ChatGPT?

What OpenAI actually says

Worth quoting precisely, because a lot of pages on this subject state things OpenAI never wrote.

Asked what types are supported, the File Uploads FAQ answers in one sentence:

All common file extensions for text files, spreadsheets, presentations, and documents.

Four families. Video is not one of them, and neither is audio. There is no published list of extensions to check, so that sentence is the whole official answer.

Which is why the honest framing is not listed among the supported types, rather than a categorical claim about what ChatGPT can and cannot process. The documentation tells you what is supported for upload. Making a broader claim from its silence would be inventing an answer, which is what most articles on this question have done.

What to do instead

The reliable path, and it works well.

1. Get a transcript

   Any transcription tool produces one. If the video is on YouTube, it very likely already has captions you can download.

2. Save it as a text file

   Plain .txt. This is the most reliably supported format there is, with nothing to go wrong.

3. Upload the transcript

   Now ChatGPT can summarise it, pull quotes from it, answer questions about it, or reshape it into something else.

For a long video, the transcript may itself be past 2 million tokens. An hour of speech runs to roughly ten thousand words, so most single videos are fine, and a series or a full day of recordings may need splitting.

If the visuals are the point

Transcription only captures what was said. If your question is about what was shown, take screenshots of the frames that matter and upload those as images.

Images are supported and capped at 20 MB each. There is also a ceiling of around 100 images per conversation, so pick the frames that carry the answer rather than sampling every few seconds.

Why the file will not simply upload

If you try, you will get either a rejection or a file ChatGPT cannot do anything useful with. Video is also large: a few minutes of high quality footage can approach the 512 MB ceiling on its own, so even a supported path would run into size before it ran into anything else.

Transcribing first solves the format question and the size question at the same time. An hour of video is hundreds of megabytes. Its transcript is a few dozen kilobytes.