Create Viral Content with Make.com & DeepSeek AI

Overview

This is an automated viral article generation solution based on DeepSeek.

Achieve personalized stylized image-text creation:

1. Topic Input - Set topic and style in Notion
2. Material Gathering - ExaAI retrieves relevant materials from the internet
3. Style Writing - DeepSeek R1 creates content imitating specified style
4. Content Structuring - GPT-4o Mini structures the article
5. Image Matching - 302AI retrieves matching images for paragraphs

From topic input to viral content generation automated process

---

Core Decision Factors

When choosing an AI writing solution, consider:

• Style Imitation - Can it precisely replicate specific author styles
• Image-Text Generation - Auto-match high-quality relevant images
• Automation Level - Workflow integration and ease of use
• Content Quality - Logic, accuracy, avoiding hallucinations
• Material Efficiency - Fast scraping and integration of internet materials

---

Technical Specifications

Specification	Value	Notes
Core AI Model	DeepSeek R1	Reasoning model, excels in Chinese
Image Retrieval	302AI	Cost-effective
Material Scraping	ExaAI	Recommend returning 4 results
Text Segmentation	GPT-4o Mini	Structured output
Automation Platform	Make.com	Routing and function handling
API Calls	TogetherAI	95%+ success rate
Image Storage	Cloudinary	Sufficient free quota
Image Strategy	1 per 3 paragraphs	Up to 6-7 images

---

Prerequisites

Before starting, ensure you have:

• Make.com account (free registration)
• Notion account and database
• DeepSeek API key (recommended via TogetherAI)
• ExaAI API key (material scraping)
• 302AI API key (image retrieval)
• Cloudinary account (image hosting)

Notion Database Structure

Create viral content database with these fields:

• Article Topic (Text) - Desired creation topic
• Writing Style (Select) - Various styles available
• Material Source (Select) - ExaAI scraping/manual input
• Status (Select) - Pending/Processing/Completed
• Generated Result (Text) - Image-rich article content

---

Writing Style Options

Multiple writing styles selectable in Notion

DeepSeek R1 supports imitating various well-known styles:

Emotional:

• High emotional resonance
• Warm and healing

Knowledge-based:

• Literary and intellectual
• Substantial depth

Hot Topics:

• Passionate and energetic
• Professional analysis

You can customize any style by simply adding style descriptions in Notion.

---

Generation Effect Showcase

Emotional Style Example

Viral article generated in emotional style

Characteristics:

• Eye-catching title, strong emotional pull
• Short sentence rhythm, strong resonance
• Clear viewpoint, sparks discussion

Intellectual Style Example

Article generated in intellectual style

Characteristics:

• Literary and intellectual expression
• In-depth with quotes and references
• Flowing narrative

---

Workflow Architecture

Phase 1: Material Acquisition

Supports two material sources:

ExaAI Auto-Scraping:

• Generate search keywords based on topic
• Scrape relevant materials from internet
• Recommend setting to return 4 results

User Manual Input:

• Paste materials directly in Notion
• Suitable for content with specific sources

Set/Get Variable implements material source switching

---

Phase 2: Stylized Writing

Use DeepSeek R1 for stylized creation:

Prompt Design:

You are a senior content creator, skilled in {{writing_style}} style.
Create a viral article based on the following materials.

Requirements:
1. Completely imitate {{writing_style}} writing characteristics
2. Eye-catching title that sparks curiosity
3. Clear logic, distinct viewpoint
4. Suitable for social media distribution

Materials: {{materials}}

---

Phase 3: Content Structuring

Use GPT-4o Mini to structure the article:

Purpose:

• Split long article into paragraph array
• Facilitate subsequent image matching
• Lower cost, high efficiency

Segmentation Strategy:

• Combine every 3 paragraphs into one group
• Use Make’s slice and join functions
• Generate image keywords for each group

---

Phase 4: Image Matching

Call 302AI to retrieve images for each paragraph group:

Process:

1. Generate keywords based on paragraph content
2. Call 302AI image search API
3. Get image URL
4. Upload to Cloudinary image hosting
5. Embed in corresponding article position

Image Quantity:

• 1 image per 3 paragraphs
• Up to 6-7 images total

---

Important Notes

Things to note during use:

1. Hallucination Issues - DeepSeek may produce inaccurate content, need human verification for factual information

2. Image Copyright - Images retrieved from internet have copyright risks, users assume responsibility

3. API Stability - Recommend using TogetherAI to call DeepSeek for higher success rate

4. Not a Magic Tool - Viral success depends on trending topics, algorithms, audience, etc.

5. Configuration Complexity - Initial setup involves multiple APIs and complex module connections

---

Use Cases

Recommended For

• Content Creators - Need to quickly batch generate diverse styled image-text content
• Marketing Teams - Customize content marketing for different audiences
• Chinese AI Writing Users - Value DeepSeek’s Chinese creation ability
• Automation Enthusiasts - Willing to learn Make advanced features

May Not Suit

• Content requiring extreme factual accuracy (like news, academic)
• Users unwilling to assume image copyright risks
• Automation beginners or those with limited time
• Users seeking “one-time solution” to viral content

---

FAQ

What writing styles are supported?

Supports multiple styles including emotional, knowledge-based, and hot topic styles. You can customize any style by adding style descriptions in Notion.

How are images matched?

The workflow generates keywords for each paragraph, calls 302AI to retrieve images, matching one image per 3 paragraphs, up to 6-7 images total.

Does DeepSeek have hallucination issues?

DeepSeek excels at style imitation but may hallucinate. We recommend human verification for factual content.

Can it batch generate on schedule?

Yes. Set scheduled runs, like every 60 minutes, to achieve batch content production.

---

Next Steps

After mastering the basics, you can try:

• Add more custom writing styles
• Integrate more image sources (like open-source libraries)
• Add content quality assessment mechanism
• Build multi-platform auto-publishing functionality

Feel free to leave comments if you have questions!