YouTube Transcripts
YouTube Transcripts gets the transcript of a YouTube video from its link: the text of what is said in the video, taken from its subtitles. You can get it as a single text or split into time segments, each with a link that opens the video at that moment. No credential is needed.
You can use it as a step in a flow, with the link in Video URL, or as a tool for an agent: connect its Tool output to the agent’s Tools input and the agent uses it when it is given a YouTube link.
When to use it
- You want to summarise training videos, talks or webinars without watching them in full.
- You need to pull out the main ideas from competitors’ videos.
- An agent has to answer questions about a video’s content.
- You want to turn a video into an article or a social media post.
Configuration
| Field | What it does |
|---|---|
| Video URL | The YouTube video link, for example https://www.youtube.com/watch?v=... or https://youtu.be/.... |
| Transcript Format | How the transcript is delivered: text (a single text) or chunks (time segments). Default: text. |
| Language | The subtitle languages you want, in order of preference and separated by commas, for example en,es. When empty, English subtitles are used. |
Processing a list
| Field | What it does |
|---|---|
| Batch Input | A list of items, for example video links. One transcript is fetched for each. |
| Batch Mode | When on and there is a list in Batch Input, each item is processed and all results come out together in one list. On by default. |
| Parallel Executions | How many items from the list are processed at the same time, from 1 to 5. Default: 3. |
| Skip if Empty Batch Input | When on and the list arrives empty, the component doesn’t run and neither do the following steps. On by default. |
| Skip Branch if Empty Output | When on and nothing is returned, the following steps don’t run. On by default. |
Advanced options
| Field | What it does |
|---|---|
| Chunk Size (seconds) | Only with chunks. The length of each segment in seconds. Default: 60. |
| Tool Name | The name the agent sees for the tool. |
| Tool Description | What the agent reads to know when to use it. |
| Tool response mode | How the result is handed to the agent. Leave Standard. |
Output
- Data: with
text, one item with the whole transcript. Withchunks, one item per segment, with its text, the second it starts and ends, and a link to the video at that moment. If the link isn’t valid or the video has no subtitles in the requested languages, the result says so. - Tool: the transcript ready for an agent. Connect it to the agent’s Tools input.
Example
A marketing team wants a summary of every webinar it publishes on YouTube.
- Add YouTube Transcripts and paste the webinar link in Video URL.
- Type
enin Language and leave Transcript Format attext. - Connect the Data output to an agent instructed to summarise the transcript in five key points.
- If you want to quote specific moments, switch to
chunks: each idea in the summary can link to the exact minute of the video.
Good practices
- Always set the video’s language in Language: if you leave it empty and the video has no English subtitles, no transcript is returned.
- It only works with videos that have subtitles, whether added by the owner or generated automatically by YouTube.
Related components
- Video: details of a YouTube video.