Skip to main content

How to Summarize YouTube Transcripts with AI Using Aisle Workflows

Alex Blom
,

Today's guide shows how to use Aisle Workflows, combined with tools like Supadata, to extract and summarize YouTube transcripts with AI.

This is useful when you're doing market research and need to process multiple videos quickly. Or when you're looking for specific mentions of topics across content. Or when you have an AI startup to promote—and want to learn from some of the classic CS183 lectures.

The sample workflow pulls the transcript, runs it through an LLM to summarize or extract what you need, and optionally stores it for later reference. Once built, you can run it on any video. Your team can too.

What you need:

Connect Supadata

Go to Connectors in your Aisle workspace. Scroll down to find Supadata and click Connect. You'll need a Supadata API key. If you don't have one, sign up here and grab your key from their dashboard.

Once connected, Supadata makes several tools available as workflow nodes:

  • Scrape webpage content
  • Get YouTube video transcripts
  • Pull all videos from a YouTube channel

Many integrations in Aisle work this way—connect once, get access to multiple tools you can use across different workflows.

Supadata connection screen

Create a New Workflow

Go to Workflows and click New Workflow. Name it something descriptive: "YouTube Transcript Summarizer" works.

Workflows in Aisle are how you chain multiple prompts or nodes into a reusable flow. Workflows can be shared with your team via chat, run on a schedule, or even triggered by API or external webhooks.

You'll see a blank workflow canvas with a Start node already there. The Start node is important—whatever inputs you define here become the parameters people provide when they run it.

Here, I make a video_id input so we can pass the video to scrape with a default for when users forget:

  • Name: video_id
  • Type: Text
  • Default value: (Optional) Paste a YouTube video ID here for testing

Start node configuration

Add the Supadata Node

Click the + button below the Start node. Search for "Supadata" in the node library. Select "Get YouTube Transcript" from the Supadata tools.

The node appears on your canvas. Click it to open the configuration panel.

You'll see two fields:

  • API Key: Select the Supadata connection you just created
  • Video ID: Set this to be a reference variable, and select the video_id we defined earlier

Each node in a workflow is a discrete step. Nodes can reference outputs from previous nodes (reference variables), which is how you chain operations together.

Supadata node configuration

Test the Workflow

When building workflows I typically use the Run mode for testing. Click Run Workflow in the top right. Here you can watch the nodes run in sequence and validate the output—and workflow—operates as expected.

Testing as you build is important. Each time you add a node, run the workflow to make sure it works before adding the next step. This makes debugging much easier than building the whole thing and testing at the end.

Workflow running

Options From Here

Now we have the transcript—there are a lot of options. We might use the memory feature to create a new document/context to be used in other workflows and prompts later.

Or—we might add a follow-on Prompt node that summarizes the transcript from a point of view or looks for mentions. This is then a useful tool for our team.

Workflow with prompt node

Deploy to Chat

This is where workflows become really sticky—Aisle makes it easy to distribute your Prompts and Workflows in multiple ways.

Go to Entry Points in your workflow settings. Entry points are how you make workflows available. You can deploy the same workflow in multiple ways:

  • Chat: Team members run it as a command in Aisle chat
  • API: Call it from external systems
  • Schedule: Run it automatically on a schedule
  • Webhook: Trigger it from external events (i.e. Gmail received, Slack message)

We're setting up a chat entry point.

To add a Chat Entry Point: Click Add Entry Point → Chat.

Configure:

  • Name: Something your team will recognize, like "Summarize YouTube Video"
  • Who can access: Choose who on your team can run this (just you, specific people, or everyone)
  • Follow-up behavior: This defines what happens after the workflow runs and returns results to chat. You can either:
    • Run the workflow again (if someone provides another video)
    • Switch to conversational mode with a model (so they can ask follow-up questions about the summary)

Using It in Chat

Now you and your team can run this workflow directly from chat. On the chat launcher—just search for our workflow. You will notice users see a form to fill out the video_id. What we have done here is build a useful tool that then was no effort to deploy to our team.

Chat launcher demo

The same workflow could also be deployed as an API endpoint, run on a schedule, or triggered by a webhook—without changing any of the workflow logic.