Can Claude watch YouTube videos?

Paul By Paul, who built Transcript Any Video
Published

No. Paste a YouTube link into Claude and it gets a web address, not the video, and Claude doesn't accept video or audio files either. What it can work with is the transcript, as text: pasted by you, read off the page by Claude in Chrome, or made on your computer by a Claude Code tool.

Below: what Claude does with a link, the free method, the two Claude add-ons that get closer, and what to do when the video has no captions. ChatGPT's options are not the same, they're covered in Can ChatGPT watch YouTube videos?

What does Claude do with a YouTube link?

Anthropic's help lists what you can upload to Claude. Documents: PDF, DOCX, CSV, TXT, HTML, ODT, RTF, EPUB, JSON and XLSX. Images: JPEG, PNG, GIF and WebP. No audio format and no video format are on the list.

Even motion inside an image is dropped. Anthropic's vision documentation says animations are unsupported, and “only the first frame is used.”

A link is just text. The developer of claudetube, an open-source tool for Claude Code, puts it bluntly in his README: when you share a YouTube link, Claude “sees nothing” but the URL.

So if Claude answers about a video from the link alone, test it. Ask what the speaker says at a precise minute. An answer built from the title, or from a page written about the video, won't hold up.

How do you get Claude to summarize a YouTube video?

You give it the words. YouTube's help says you can see the full transcript of any video that has captions. On a computer:

  1. Open the video on YouTube.
  2. In the video description, click Show transcript.
  3. Select the whole transcript and copy it.
  4. Paste it into Claude, then ask: a summary, the steps the speaker gave, a quiz on the lesson.
  5. For a long video, save the text as a .txt file and attach it. Claude takes up to 20 files per chat, 500 MB each.
  6. For a course you'll keep coming back to, add the file to a Claude Project. Project files stay available across conversations.

This is the method in the top tutorial on this query, a 2024 video by Evan Manza. As he says while doing it, Claude summarizes “based off the transcript because it can't actually access the video itself.”

Can Claude in Chrome watch a YouTube video?

Claude in Chrome is Anthropic's browser extension. Its help says it lets Claude “read, click, and navigate websites alongside you”, from a side panel next to the page. Two conditions: a paid plan (Pro, Max, Team or Enterprise), and Google Chrome itself. Anthropic says other Chromium-based browsers and mobile devices aren't supported.

What it reads is the page. Anthropic describes reading, clicking, typing, navigating and filling forms. It describes no listening. On a YouTube page, that leaves the title, the description and the text on screen, such as the transcript panel once you open it. A video with no captions has no transcript panel to read.

One doubt comes up in Claude user groups. A member of a Claude community group on Facebook asked whether Claude in the browser reads the full transcript or only part of it. A quick check: ask Claude what's said in the last two minutes. If it can't tell you, paste the full transcript into a chat instead.

Can Claude Code watch YouTube videos?

Closer, yes, through tools built by developers, not by Anthropic. Claude Code runs commands on your computer, so it can drive the free tools that do the work: yt-dlp to download the video, ffmpeg to cut frames, Whisper to turn speech into text.

Two limits. It's a terminal setup, not a button. And frames are images, which is costly: one commenter on the Hacker News thread called it a “pretty terribly expensive way to watch a video with Claude.” If you need the slides or the code on screen, frames earn their cost. If you need what was said, the transcript alone is enough.

Downloading someone else's video also raises YouTube's terms. They're summed up in the guide on how to get a transcript of a YouTube video without captions.

Is there an assistant that takes the link itself?

Google documents one for Gemini. Its help says you can ask Gemini about YouTube videos, and that it “uses public information from YouTube” to answer. The authors of both Claude tools above also name Gemini as the model that takes video.

What do the top results leave out?

Today, the results for this query are a Hacker News thread, a 2024 tutorial and forum posts. No guide among them. The tutorial covers copy-paste and stops there: it never says what to do when Show transcript isn't there. The Hacker News tool solves that, but only for people at ease in a terminal. Neither says that Claude takes no video files at all, or that Claude in Chrome needs a paid plan and Chrome itself.

What if the video has no captions?

Then the chat method and Claude in Chrome hit the same wall: no captions, no Show transcript, no text. Only the Claude Code tools get past it, because they fall back to Whisper on the audio.

Without Claude Code, free options exist: your computer's live captions, or Whisper running on your own machine. Both are in the no-captions guide, with their limits. If the transcript button is missing on a video that does have captions, see YouTube transcript not showing.

No captions, no terminal, and you need the text in Claude today? That's the case I built Transcript Any Video for. You paste the YouTube link, and it transcribes the video's audio, not its captions, so it works on videos that have none. You get it word for word, with just your email, no password. Videos can run up to 2 hours. You buy credits once, 10 for $1.99, no subscription, and you get a refund if the transcript is wrong. Then paste the text into Claude.

Transcribe a video

FAQ

Can Claude transcribe a YouTube video?

Not in the Claude app. It takes no audio uploads, so there's no sound for it to transcribe. In Claude Code, the transcribing is done by Whisper, a separate speech-to-text model running on your computer. Claude then reads what Whisper wrote.

Can I use Claude with YouTube on my phone?

Anthropic says Claude in Chrome isn't supported on mobile devices. On a phone, what's left is pasting a transcript into the Claude app. YouTube's help adds that tapping any line of the transcript jumps to that moment, which is handy to check a point Claude picked out.

How long a video can Claude handle?

What counts is the length of the transcript, not of the video. Anthropic's help says project files must fit within Claude's context window, and that token limits may apply to long extracted text. For a long lecture, split the transcript by chapter and ask about one part at a time.

How do I stop Claude from inventing parts of the video?

Ask for quotes first. Anthropic's guide on reducing hallucinations suggests having Claude pull exact quotes before it answers, and say “No relevant quotes found” when there are none. With a transcript, every point of the summary then rests on a line you can find.

Sources