Automation Workflows with AI

Automated viral video pipeline using n8n and DeepSeek 3.1 with free TTS, captions, music, and publishing.

Overview

autom8n automates the creation and publishing of shortform video content. Using n8n as the orchestrator and DeepSeek 3.1 to write scripts, the pipeline generates voiceovers, visuals, captions, and music — and can even auto-upload.

Features

Script generation using DeepSeek 3.1
Text-to-speech with local TTS engines (e.g., Piper, Bark, Coqui)
Automatic captioning using Whisper or WhisperX
Background music via free audio sources or generated loops
Video rendering with FFmpeg or Auto-Editor
Optional uploading/posting to TikTok, YouTube Shorts, etc.
100% free and unlimited with self-hosted components

Architecture

Trigger: Scheduled or event-driven n8n workflow
Script Generation: DeepSeek call to generate a short, viral script
TTS: Convert script to audio using a local TTS engine
Music: Add royalty-free or auto-generated background music
Captioning: Generate subtitle timing and overlay with Whisper
Rendering: Merge visuals, audio, and captions using FFmpeg
Publishing: Upload via APIs or headless browser automation

Requirements

Node.js
Docker (optional, for Whisper or n8n)
Python (for Whisper/WhisperX)
n8n (self-hosted or desktop)
DeepSeek 3.1 (local model or API)
FFmpeg
A local TTS engine (Piper recommended)

Prereq's

Docker Desktop
Github
Optional NodeJS

Setup

Clone the repo:

git clone https://github.com/maxwellvolz/viraln8tion.git
cd viraln8tion

Run Services in Docker Containers

Service	Purpose	Access
n8n	Automation Tool	Console
MiniIO	Object Storage	admin :: password123 console
Kokoro TTS	Text-to-Speech	Console
Baserow	Database UI	Console

MiniIO http://host.docker.internal:9001 Kokoro TTS

NCA Toolkit http://host.docker.internal:8080

Run all 4 of these in separate Terminals (cuz they take a while):

# n8n
docker run -d --name n8n -p 5678:5678 -e WEBHOOK_URL=http://host.docker.internal:5678 -e N8N_DEFAULT_BINARY_DATA_MODE=filesystem -v C:\Docker\n8n-data:/home/node/.n8n docker.n8n.io/n8nio/n8n
 
# miniIO
docker run -p 9000:9000 -p 9001:9001 --name miniio -v C:\Docker\minio-data:/data -e MINIO_ROOT_USER=admin -e MINIO_ROOT_PASSWORD=password123 quay.io/minio/minio:RELEASE.2025-04-22T22-12-26Z server /data --console-address ":9001"
 
# Kokoro TTS
docker run -d --gpus all -p 8880:8880 --name kokoro-tts ghcr.io/remsky/kokoro-fastapi-gpu:v0.2.2
 
# Only if ur hardware sucks - Kokoro TTS CPU Edition
docker run -p 8880:8880 --name kokoro-tts-cpu ghcr.io/remsky/kokoro-fastapi-cpu:v0.2.2
 
# Baserow
docker run -d --name baserow -e BASEROW_PUBLIC_URL=http://host.docker.internal:85 -v C:\Docker\baserow-data:/baserow/data -p 85:80 -p 443:443 --restart unless-stopped --shm-size=256mb baserow/baserow:1.32.5

Activate n8n key - from email

Verify Setups

Check Docker Desktop or the Consoles listed above.

n8n
1. Make Account
MiniIO
1. Make Account
2. Make Database nca-toolkit
3. Generate Access and Secret Keys using the default root credentials
4. Save these for later
Kokoro
Baserow
1. Create Workspace
2. Create Database TikTok
  1. Add Tables Videos and Scenes

Database Table Setup - EZ Mode

Save a .csv
Import as View for Videos and Scenes

Videos

id,Title,Description,Script,Final Video URL,Video + Captions URL,Video + Audio URL,Raw Video URL,TTS Audio,TTS Voice,Scenes,Captions URL,Generative Style,Initial Prompt,Status,Image Provider
1,,,,,,,,,,,,,,,

Scenes

id,Record ID,Prompt,Duration,Image,Video Clip URL,Videos,Image Provider
1,1,,0.00,,,,

Install NCA Toolkit

Service	Purpose	Access
NCA Toolkit	Video Editing API	Console

Usage:

images to videos
combine clips
transcribe clips

Replace your_access_key and your_secret_key with the values from MiniIO:

docker run -d -p 8080:8080 --name nca-toolkit -e API_KEY=thekey -e S3_ENDPOINT_URL=http://host.docker.internal:9000 -e S3_ACCESS_KEY=your_access_key -e S3_SECRET_KEY=your_secret_key -e S3_BUCKET_NAME=nca-toolkit -e S3_REGION=None stephengpope/no-code-architects-toolkit:latest

OpenRouter

Make Account
Copy Api Key
Add to n8n

n8n Automation

Import JSON
Duplicate
Disconnect all but first lane
Double-click On form submission and open the Test URL

http://host.docker.internal:85/api-docs/database/163

then Settings

amazing facts about linus torvalds

n8n Setup

Node → HTTP Request

Method: Post
Auth: None
Send Headers: True
- Name: Authorization
- Value: Token enter_your_baserow_api_token
Send Body: True
- Name: Title, Value: {value}

Rate Limited? Local LLM Setup

Prereq's

Python

Make /local_llm/.env with: HF_TOKEN=your_huggingface_key

Setup

cd local_llm
.venv\Scripts\activate
pip install -r requirements.txt

Run it

cd local_llm
make tinyllama          # Spin up tinyllama on port 8000
make tinyllama-down     # Stop tinyllama container
make deepseekcode       # Spin up DeepSeekCode on port 8000
make deepseekcode-down  # Stop DeepSeekCode container
make mistral            # Spin up Mistral 7B on port 8000
make mistral-down       # Stop Mistral container
make stop-all           # Kill all running containers

Postman - Test LLM Endpoint

Type: POST
URL: http://localhost:8000/v1/chat/completions

Headers

Authorization | Bearer Test
Content-Type | application/json

Body

Model Names: deepseek-ai/deepseek-coder-1.3b-instruct, mistralai/Mistral-7B-Instruct-v0.1, TinyLlama/TinyLlama-1.1B-Chat-v1.0

raw:

{  
   "model": "deepseek-ai/deepseek-coder-1.3b-instruct",  
   "messages": 
   [    
      {      
      "role": "user",      
      "content": "solve the fibonacci sequence with python using a list comp"    
      }  
   ]
}

Youtube Credentials

To automate posting to YouTube using n8n, you need to use the YouTube node (which uses the YouTube Data API v3). Here's a full setup that lets you upload videos programmatically.

✅ Requirements

Google Cloud project
YouTube Data API v3 enabled
OAuth2 credentials
n8n YouTube node configured

🔧 Setup Guide

1. Create OAuth Credentials

Go to: https://console.cloud.google.com/
Create a project or use existing one.
Enable YouTube Data API v3
Go to APIs & Services → Credentials

In Google Cloud → Credentials
Click Create Credentials → OAuth Client ID
Choose:
```
Application Type: Web application
```
Set name: n8n YouTube Uploader

Add Authorized Redirect URI:

http://localhost:5678/rest/oauth2-credential/callback

Save new Client ID and Client Secret.

n8n credentials

Go to Credentials → New → YouTube OAuth2
Use the new Web Client ID + Secret
It will now redirect correctly and allow you to sign in

Stable Diffusion Locally

Pre-reqs

mkdir models outputs

docker-compose.yml

services:
   ...
  stable-diffusion:
    image: siutin/stable-diffusion-webui-docker:latest-cuda-12.1.1
    container_name: sd
    command: ["bash", "webui.sh", "--api", "--listen"]
    ports:
      - "7860:7860"
    deploy:
      resources:
        reservations:
          devices:
            - capabilities: [gpu]
    volumes:
      - ./models:/app/stable-diffusion-webui/models
      - ./outputs:/app/stable-diffusion-webui/outputs
    restart: unless-stopped

Run Service

docker compose up -d stable-diffusion
# wait 2 minutes

Test Image

Note: Stable Diffusion WebUI API (AUTOMATIC1111-based) returns a base64-encoded PNG, not a raw image file. We can convert to a .png to view the result.

curl http://localhost:7860/sdapi/v1/txt2img \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "A cyberpunk-styled close-up of a grieving parent'\''s hands cradling a jar filled with dirt instead of ashes, illuminated by a flickering neon sign outside a dimly lit funeral home. The jar reflects eerie blue and purple hues, casting long shadows on the teardrop-streaked face in the background.",
    "negative_prompt": "blurry, ugly, deformed, bad anatomy, low quality, abstract, cropped, out of frame, messy, bad lighting, text, watermark, distorted, wrong perspective",
    "steps": 20,
    "width": 576,
    "height": 1024
  }' | jq -r '.images[0]' | base64 -d > result.png

Upgrade to `SD-XL` model

Download

Place in /models

curl http://localhost:7860/sdapi/v1/txt2img \
  -H "Content-Type: application/json" \
  -d '{
    "sd_model_checkpoint": "sdxl_base_1.0.safetensors",
    "prompt": "A cyberpunk-styled close-up of a grieving parent'\''s hands cradling a jar filled with dirt instead of ashes, illuminated by a flickering neon sign outside a dimly lit funeral home. The jar reflects eerie blue and purple hues, casting long shadows on the teardrop-streaked face in the background.",
    "negative_prompt": "blurry, ugly, deformed, bad anatomy, low quality, abstract, cropped, out of frame, messy, bad lighting, text, watermark, distorted, wrong perspective",
    "steps": 20,
    "width": 576,
    "height": 1024
  }' | jq -r '.images[0]' | base64 -d > result.png

Automation Workflows with AI

Overview

Features

Architecture

Requirements

Prereq's

Setup

Run Services in Docker Containers

Verify Setups

Database Table Setup - EZ Mode

Install NCA Toolkit

OpenRouter

n8n Automation

n8n Setup

Node → HTTP Request

Rate Limited? Local LLM Setup

Prereq's

Setup

Run it

Postman - Test LLM Endpoint

Youtube Credentials

✅ Requirements

🔧 Setup Guide

1. Create OAuth Credentials

n8n credentials

Stable Diffusion Locally

Pre-reqs

Run Service

Test Image

Upgrade to SD-XL model

Upgrade to `SD-XL` model