n8n YouTube transcript not working: each error and its fix

Two rules from n8n's own docs decide this. The Code node page says: "If you use n8n Cloud, you can't import external npm modules." The community nodes page says: "Unverified community nodes aren't available on n8n cloud and require self-hosting n8n."

So on Cloud a Code node cannot load youtube-transcript, and the unverified transcript nodes on npm cannot be installed. What is left is an HTTP Request node pointed at a transcript API, or a verified community node from the nodes panel.

On a default self-hosted n8n 2.47.5, this Code node:

const { YoutubeTranscript } = require('youtube-transcript');
const r = await YoutubeTranscript.fetchTranscript('aircAruvnKk', { lang: 'en' });
return [{ json: { segments: r.length, first: r[0].text } }];

failed with Module 'youtube-transcript' is disallowed [line 1].

Two things fix it, and you need both. Install the package where n8n can find it. We installed it with npm in the same folder as n8n. For Docker, the route people on the n8n forum report working is a custom image built from n8nio/n8n with the package installed globally. Then start n8n with:

NODE_FUNCTION_ALLOW_EXTERNAL=youtube-transcript

With that set, the module loaded on our install and the code ran as far as its request to YouTube. Our test machine sends traffic through a proxy, and the Code node's runner did not use it, even with Node's proxy flags set, so that request failed with fetch failed. If your n8n sits behind a corporate proxy, you may see the same. We ran the rest of the tests in plain Node 24 with the same package version.

youtube-transcript 1.2.1 failed on 3 of 3 videos with TypeError: Cannot read properties of undefined (reading 'slice'). Version 1.3.1 returned all three.

This matters for community nodes. n8n-nodes-youtube-transcript-api 2.4.3 depends on youtube-transcript ^1.2.1. A fresh install today resolves to 1.3.1, but an install from early 2025 may still hold the old one. Reinstall the node to pick up the new version.

With no language set, the package takes the first caption track YouTube lists. That is not always the spoken language.

Two of the English videos we tried came back in Arabic. The package's code takes the first track in YouTube's list, and for those two videos the first track was Arabic. With { lang: 'en' } both returned English (286 and 341 segments).

The n8n-nodes-youtube-transcript-api node has a Language field, and it is empty by default. Fill it in.

The package compares your code with each track's code and does no fallback. On the Rick Astley video:

• en worked.
• en-US failed: No transcripts are available in en-US this video (dQw4w9WgXcQ). Available languages: en, en, de-DE, ja, pt-BR, es-419
• de failed with the same message, although de-DE is in the list.

The error prints the codes the video has. Copy one from there.

The package throws [YoutubeTranscript] 🚨 Transcript is disabled on this video whenever YouTube's answer holds no caption tracks. We got it for three different cases:

• a live stream (jfKfPfyJRdk)
• a video ID that does not exist (xxxxxxxxxxx)
• a second made-up ID (aaaaaaaaaaa)

None of those is a video whose owner turned captions off. So before you trust the message, open the video in a browser and look for the captions button. If the video has captions there and your workflow still says disabled, the answer YouTube gave your server was different from the one it gave your browser. We could not reproduce that case from a home connection, so we cannot tell you what a blocked server sees. Our RequestBlocked guide covers how YouTube treats server addresses for the Python library.

On 1.3.1 each segment is { text, duration, offset, lang }, with offset and duration in milliseconds. Text keeps the line breaks from the caption file, so join segments with a space and replace \n if you want one paragraph. We found no HTML entities such as ' in the seven videos that returned.

The Puppeteer nodes need a Chromium that the n8n container can launch. The first node's author said on the n8n forum that the browser route is resource heavy and unreliable at scale. We did not run the Puppeteer nodes.

The other way out is to stop fetching from your n8n server. A transcript API does the fetch on its own network, and n8n only makes one HTTPS call.

TranscriptFetch has one, which is our product, so weigh this section with that in mind. The TranscriptFetch n8n node is a verified community node, so it installs from the nodes panel on n8n Cloud and on self-hosted: open a workflow, click +, search for TranscriptFetch, click Install. Add a TranscriptFetch API credential with a key from your dashboard, and n8n checks the key when you save.

The node gets one transcript, a batch of up to 50, a channel's or playlist's videos, and search results, for YouTube, TikTok and Instagram. A trigger starts a workflow when a channel posts a new video. Videos with no captions can go to AI transcription, which answers the real Transcript is disabled case. There is a ready workflow to fetch one transcript you can import.

You start with 50 free credits with no card, a caption transcript costs 1 credit, and failed requests are free. We did not run the node for this post, because the test had no API key. The steps above are from its docs.