Video transcript search and Q&A with VLM Run, GPT-4 & Google Workspace

Automate video transcription and Q&A with async VLM processing that scales from short clips to long recordings.

What this workflow does

1. Monitors Google Drive for new files in a specific folder and grabs the file ID on create
2. Automatically downloads the binary to hand off for processing
3. Sends the video to VLM Run for async transcription with a callback URL that posts results back to n8n
4. Receives the transcript JSON via Webhook and appends a row in Google Sheets with the video identifier and transcript data
5. Enables chat Q&A through the Chat Trigger + AI Agent. The agent fetches relevant rows from Sheets and answers only from those segments using the connected chat model

Setup

Prerequisites: Google Drive and Google Sheets accounts, VLM Run API credentials, OpenAI (or another supported) chat model credentials, n8n instance.

Install the verified VLM Run node by searching for VLM Run in the nodes list, then click Install. You can also confirm on npm if needed. After install, it integrates directly for robust async transcription.

Quick Setup:

1. Google Drive folder watch

   • Add Google Drive Trigger and choose Specific folder. Set polling to every minute, event to File Created. Connect Drive OAuth2.

2. Download the new file

   • Add Google Drive node with Download. Map {{$json.id}} and save the binary as data.

3. Async transcription with VLM Run

   • Add VLM Run node. Operation: video. Domain: video.transcription. Enable Process Asynchronously and set Callback URL to your Webhook path (for example /transcript-video). Add your VLM Run API key.

4. Webhook to receive results

   • Add Webhook node with method POST and path /transcript-video. This is the endpoint VLM Run calls when the job completes. Use When Last Node Finishes or respond via a Respond node if you prefer.

5. Append to Google Sheets

   • Add Google Sheets node with Append. Point to your spreadsheet and sheet. Map:

     • Video Name → the video identifier from the webhook payload
     • Data → the transcript text or JSON from the webhook payload

   • Connect Google Sheets OAuth2.

6. Chat entry point and Agent

   • Add Chat Trigger to receive user questions.

   • Add AI Agent and connect:

     • a Chat Model (for example OpenAI Chat Model)
     • the Google Sheets Tool to read relevant rows

   • In the Agent system message, instruct:

     • Use the Sheets tool to fetch transcript rows matching the question
     • Answer only from those rows
     • Cite or reference row context as needed

7. Test and activate

   • Upload a sample video to the watched Drive folder.
   • Wait for the callback to populate your sheet.
   • Ask a question through the Chat Trigger and confirm the agent quotes only from the retrieved rows.
   • Activate your template and let it automate the task.

How to take this further

• Team memory: Ask “What did we decide on pricing last week?” and get the exact clip and answer.
• Study helper: Drop classes in, then ask for key points or formulas by topic.
• Customer FAQ builder: Turn real support calls into answers your team can reuse.
• Podcast highlights: Find quotes, tips, and standout moments from each episode.
• Meeting catch-up: Get decisions and action items from any recording, fast.
• Marketing snippets: Pull short, social-ready lines from long demos or webinars.
• Team learning hub: Grow a searchable video brain that remembers everything.

This workflow uses the VLM Run node for scalable, async video transcription and the AI Agent for grounded Q&A from Sheets, giving you a durable pipeline from upload to searchable answers with minimal upkeep.