YouTube transcript search in practice.

Play a quote timeline

Hear five conversations about vibe coding. Each excerpt opens the original recording near the selected sentence. You can explore the saved results without an API key.

Get the Python worksheet and sample CSV or open it in Google Colab to export the passages and timestamps as a CSV.

Make video quote cards with Remotion

Watch an 18-second example, then render it yourself from saved passages. The example includes the source links and timestamps for each card.

Add transcript search to an AI SDK agent

Install the TypeScript adapter and give your agent tools to search YouTube and podcast transcripts and look up people by name.

Export a transcript as WebVTT

Turn a saved transcript into a caption file for your video or podcast. The local Python example keeps each cue’s timing and any speaker labels in the response.

Install the YouTube transcript search SDK

For TypeScript or JavaScript, run npm install arcmira. The npm package includes the SDK and CLI. Run npx arcmira search --help to see the transcript search options.

For Python, run python -m pip install arcmira in your virtual environment. The PyPI package includes synchronous and asynchronous clients. The SDK guide shows how to authenticate and run a search.

Search the API reference offline

Look up a transcript endpoint, request parameter or response schema in Dash or Zeal. Search by HTTP path or operation ID, such as get_transcript.

Add the reference to Dash. In Zeal, open Docsets, choose Add Feed and paste the docset feed URL. You can also download the docset archive.

YouTube transcript search workflows for n8n, Dify and Node-RED

Start with a topic and get a shortlist of spoken passages with timestamps and source links. Choose the output you need, import the template, then connect your Arcmira API key in your workflow tool.

n8n: Make a clip shortlist in CSV

Search for a phrase such as “vibe coding” and export up to five passages. Open each timestamp to hear the speaker before choosing a clip. Download the n8n clip workflow · Import and run in n8n

n8n: Collect sources in JSON

Choose a date range and return up to three passages. Each result keeps the video title, source URL and timestamp. The output also records the date window the search covered. Download the n8n source workflow · Set up the source search

Dify: Pass transcript results to another step

Import the workflow into Dify Studio and connect the Arcmira plugin. It returns the search response as JSON, including available speaker labels and coverage notes. Download the Dify workflow · Set up the Dify plugin

Node-RED: Save a CSV and its source data

Run a topic search manually and write the passages to a local CSV with a companion JSON file. Each run replaces those two output files. Download the Node-RED flow · Configure the output directory