Analyze YouTube videos and auto-generate AI reports in Google Docs with DeepSeek

A compact n8n workflow that accepts a YouTube link or uploaded video, pulls a transcript via Supadata.ai, runs a language-model-based video analysis agent to produce a structured report, extracts a title/metadata, then creates and updates a Google Doc with the analysis. It's designed to automate transcription → analysis → document creation for fast, repeatable video reviews.

How it works

Trigger — Upload File or YouTube Link
A form trigger receives a youtube_url or an uploaded file/webhook event.

Transcription — Transcription using Supadata.ai
Calls the transcription API using the x-api-key header to retrieve the video transcript/text.

Analysis — Analyser
The transcript is passed to the Analyser LangChain agent which runs a tailored prompt (expert video analyst) and generates a plain-text report.

Metadata extraction — File Name Detector
The information extractor parses the analyser output to extract structured attributes such as the Title.

Aggregation & Merge
Merge/Aggregate nodes combine the analysis and extracted fields into a single payload.

Document Creation
Creating New File creates a Google Docs document using the extracted Title, and Updating Content in File inserts the analyser output into the document.

Optional Follow-ups
Additional nodes can forward the document link, send it to Slack, or store metadata in a database.

Quick Setup Guide 👉 Demo & Setup Video 👉 Course

Nodes of interest

Upload File or YouTube Link**
formTrigger (webhook) — Entry point for user-supplied links or files.

Transcription using Supadata.ai**
httpRequest — Fetches transcript from https://api.supadata.ai/... and requires the x-api-key header.

OpenRouter Chat Model / OpenRouter Chat Model1**
lmChatOpenRouter — Language model nodes connected to the Analyser and File Name Detector using the model
deepseek/deepseek-r1-distill-llama-70b.

Analyser**
LangChain agent node that contains the expert analysis prompt and generates a full plain-text report from the transcript.
Configuration includes hasOutputParser: true and retry enabled.

File Name Detector**
LangChain information extractor that extracts structured attributes like Title from the analysis output.

Merge / Aggregate**
Combines outputs from analysis and extraction into a single payload used for document creation.

Creating New File / Updating Content in File**
Google Docs nodes used to create and update documents using googleDocsOAuth2Api credentials.

What you’ll need (credentials)

OpenRouter account**
Used by OpenRouter Chat Model nodes. API key stored in the openRouterApi credential.

Supadata.ai API key**
Added in the HTTP header x-api-key in the transcription request.

Google Docs OAuth2**
googleDocsOAuth2Api credential used for creating and updating Google Docs.

Optional integrations**
Slack webhook, Google Drive, or database credentials if adding notifications or persistent storage.

Recommended settings & best practices

Prompt control**
Keep the Analyser prompt explicit about required sections, output style, and how to handle missing transcripts.

Retries & timeouts**
Enable retries for long-running model or HTTP calls. Configure proper HTTP request timeouts.

Rate limits**
Respect transcription and model provider rate limits. Add throttling if needed.

Input validation**
Validate the youtube_url before processing and handle transcript failures gracefully.

Chunk transcripts**
Split long transcripts into chunks before sending to the LLM to avoid context limit issues.

Logging & audit**
Store transcripts, analysis results, and metadata for debugging and traceability.

Security**
Store API keys as n8n credentials rather than plaintext.

Document naming**
Sanitize the extracted Title to prevent invalid filename characters.

Monitoring**
Add error notifications via email or Slack for failed runs.

Customization ideas

Alternative transcription providers**
Replace Supadata.ai with AssemblyAI, Whisper (self-hosted), or YouTube captions.

Multiple output formats**
Export results to Google Docs, PDF, or JSON metadata.

Speaker diarization**
Include speaker labels and timestamps in the analysis.

Summaries & highlights**
Add TL;DR summaries and timestamped key moments.

Content classification**
Use additional LLM nodes to detect sentiment, category, or compliance issues.

Thumbnail generation**
Capture frames from the video to generate thumbnails.

Webhook callbacks**
Send the document link to Slack, email, or other systems.

Model routing**
Use smaller models for short videos and higher-quality models for long videos.

Human review pipeline**
Create a review queue for manual verification before publishing results.

Tags

video-analysis
transcription
n8n
langchain
automations
google-docs
openrouter
supadata
reporting
workflow