Partly. ChatGPT accepts video files, even on Free, but may skip parts or the audio. With a YouTube link it reads pages and transcripts, not the frames.
By HoverNotes Team•9 min read
Partly. ChatGPT accepts video files as attachments, including on the Free plan, but OpenAI says its analysis can be incomplete: it may not cover the whole video or get the audio right. Paste a YouTube link and it usually reads the video's page and a transcript found on the web, not the frames. Code, slides and diagrams that nobody reads out loud are what it is most likely to miss.
Search results disagree on this. Some pages say ChatGPT can't watch videos at all; others say it can analyze anything you upload. Both are out of date. The facts below come from OpenAI's own help center and pricing page, checked on October 30, 2026, plus a test we ran the same day.
Video uploads work on Free and paid plans. Availability "varies by platform, upload method, and account". If your phone's photo picker won't let you select a video, try attaching it as a file instead.
The analysis is best-effort. ChatGPT "may use tools to analyze uploaded videos, but its analysis can be incomplete or inaccurate. It may not analyze the entire video or accurately interpret its audio."
Videos count toward your file-upload limits. According to Uploading files and audio to ChatGPT, that's up to 80 files every 3 hours, and 3 uploads a day on the Free plan. OpenAI doesn't publish a separate size or length limit for video files.
In practice, ChatGPT handles a video by working on pieces of it: still frames and, when it can, the audio. That's fine for a short clip ("what happens in this video?", "write a caption for this"). For an hour-long lecture, expect it to sample rather than watch every second, and ask it which parts it actually looked at.
Two related features are easy to confuse with video uploads:
Feature
What it does with video
Plans and platforms
Video file upload
Analyzes an attached file, possibly incompletely
Free and paid; counts toward upload limits
Audio file upload
Transcribes and discusses an audio file. MP4 or WebM files recognized as video are rejected
Paid plans only
Voice with video or screen sharing
Lets ChatGPT see your camera or phone screen during a voice chat
Paid plans, iOS and Android, using the Advanced voice option (ChatGPT Voice)
Record
Transcribes audio, such as a meeting or voice note, into notes
Plus, Pro, Business, Enterprise and Edu, in the macOS app only (ChatGPT Record)
Record hears audio but doesn't see the screen. Screen sharing in voice mode is for a live conversation on your phone, not for making notes on a 40-minute lecture.
ChatGPT can search the web on every plan. Nothing in OpenAI's help center says it can stream a YouTube video. What it does with a link is look for text about the video: the YouTube page, the creator's site, and transcripts published on third-party sites.
We tested it on October 30, 2026. We pasted the link to 3Blue1Brown's "But what is a neural network?" lecture and asked ChatGPT to summarize it and describe what's on screen at 3:00. It returned a good, timestamped summary, citing a third-party transcript page and the lesson page on 3blue1brown.com. For the 3:00 question, it described the scene from the narration and added: "I verified the timestamp against an available transcript, but I couldn't independently inspect the exact video frame."
That's the pattern to expect:
Popular videos with captions: a useful summary, built from the transcript and pages about the video.
New, obscure or caption-less videos: ChatGPT may find nothing, ask you for the transcript, or summarize the title and description. Check its sources before trusting a summary.
Private, unlisted or paid-course videos: web search can't reach them.
OpenAI's built-in browser guide describes a ChatGPT browser extension for Chrome: mention @Chrome or pick a tab, and ChatGPT can work with your open tabs and signed-in sessions. Paid plans other than Go also get a cloud browser in ChatGPT Work. OpenAI presents both for web tasks such as research, bookings and forms. Neither is described as watching a video while it plays. (ChatGPT Atlas, OpenAI's browser, has been retired.)
A transcript is only the words. In a lot of learning videos, the important part is on screen:
Code in an editor or terminal that the instructor types without reading out.
Diagrams and whiteboard work, such as a network topology, an anatomy sketch or a system design.
Slides with formulas, tables or definitions the speaker summarizes in a sentence.
Demos: which menu was clicked, which setting changed, what the output looked like.
When ChatGPT works from a YouTube link, it never sees any of that. When it works from an uploaded file, it may see some frames, but OpenAI's own documentation warns it may not analyze the entire video. And ChatGPT's notes come with no timestamps you can click to go back to the moment and check.
This works on any plan, for YouTube videos and anything else with captions.
Get the transcript. On YouTube, open the description, click Show transcript, and copy it. For other options, see how to transcribe a YouTube video.
Paste it with a specific prompt. For example: "Here's a lecture transcript with timestamps. Write study notes with headings, key definitions and a timestamp for each section. Mark anything that sounds like it refers to something on screen ('as you can see here', 'this line')."
Capture the visuals yourself. Pause on each slide, diagram or block of code and take a screenshot (Win+Shift+S on Windows, Cmd+Shift+4 on a Mac). Attach the important ones and ask ChatGPT to transcribe the code or explain the diagram. Remember that images count toward the Free plan's 3 uploads a day.
Use a file upload for short clips. If you have the video file and the right to use it, attach it and ask for a summary. Then ask, "Which parts of the video did you analyze?" and spot-check a few moments.
Check before you study from it. Open the sources ChatGPT cites, compare a couple of sections with the video, and fix anything it got wrong.
Once the notes are right, ChatGPT is very good at the next step: explaining a confusing part, quizzing you, or turning your notes into practice questions.
The workflow above has two weak spots: you're copying and pasting while the video plays, and the visuals depend on you remembering to screenshot. If you learn mostly from video, a tool that runs beside the player saves both steps.
HoverNotes is our product, so weigh this accordingly. It's a Chrome extension that opens a notes panel next to any video in your browser, including YouTube, Udemy, Coursera and university portals you're signed in to. AI Notes come from our own custom-trained models and are linked to timestamps as the video plays. Press Ctrl/⌘+K in the HoverNotes panel to save the frame on screen as a timestamped screenshot, so the code and diagrams a transcript misses end up in your notes. Everything saves as Markdown to your Obsidian vault or a folder you choose. The free plan includes 20 minutes of AI Notes with no card; Unlimited Pro is $18 a month or $120 a year in the US. It's built for learning from videos, not for meeting notes. There's more on the YouTube side in YouTube to notes, and a side-by-side in HoverNotes vs ChatGPT.
Many people use both: a watch-along tool to capture the video, and ChatGPT to understand and review what they captured.
Not directly. Paste a YouTube link and ChatGPT searches for text about the video, such as its page and transcripts published on other sites, and works from that. In our October 2026 test it said it couldn't inspect the actual video frame. To have it look at the picture, you need to upload the video file or screenshots.
Yes, when it can get the words. It works best when you paste the transcript yourself. With just a link, it can summarize popular videos that have transcripts online, but it may fail or guess on new, unlisted or caption-less videos. Check the sources it cites.
Yes. OpenAI says video files can be attached on the Free plan, subject to its upload limits: Free users get 3 file uploads a day. Its analysis may be incomplete or miss the audio.
OpenAI doesn't publish a length or size limit for video files. It does say ChatGPT may not analyze the entire video, so for long lectures, use the transcript for the words and screenshots for the visuals.
Not from a video file. Audio uploads, which ChatGPT can transcribe, are on paid plans only, and MP4 or WebM files recognized as video are rejected. Save the audio track as an MP3 or M4A first, or start from the video's own captions.
On iOS and Android, paid subscribers can share their camera or screen in a voice conversation using the Advanced voice option. That's a live chat, not note-taking, and it doesn't produce timestamped notes of the video.
Sources: OpenAI help center articles linked above and chatgpt.com/pricing, checked October 30, 2026. ChatGPT changes often; if something here is out of date, tell us at hovernotes.io.