Telegram Sora AI Agent - System for generating and publishing video ads on TikTok

‍

System description

N8n automation with an AI agent on Telegram that generates promotional game videos via Sora 2 based on uploaded photos of characters from games. Once the video is created, the system offers automatic posting on TikTok.

How it works:

1. Send a character photo to (for example, from a game) on Telegram
2. AI analyzes the image and creates a detailed scenario
3. Sora 2 generates a 12-second ad video for a character
4. The bot offers to post on TikTok
5. If you agree, the video is automatically published

MAIN WORKFLOW

System components

Telegram Trigger

• Feature: Monitoring incoming messages and files
• Webhook ID: 6ffbdde8-db19-4657-b41c-8395dce7ac68
• Processes: Photos, documents, text

AI Agent - Central Coordinator

• Model: GPT-4.1-mini
• Memory: Simple Memory with a link to chat.id
• System prompt:
  
  • When received, file_id sends “Material in processing...”
  • Runs Gen_video with file_id
  • Offers a post on TikTok
  • Launches Auto_Posting if you agree

Agent tools:

send_text - Send messages to Telegram

• Chat ID from the dialog context
• Processing statuses and results

Gen_video - Generate video

• Workflow ID: create workflow Gen_video separately and select from the list in this field
• Input: image file_id
• Output: WebContentLink video in Google Drive

Auto_Posting - Posting on TikTok

• Workflow ID: Create workflow Auto_Posting separately and select from the list in this field
• Input: WebContentLink video
• Automated publishing

Simple Memory

• Saves the history of conversations
• Personalization for every user
• Session Key: chat.id

SUB-WORKFLOW: GEN_VIDEO

Video generation process

1. When Executed by Another Workflow

Getting file_id from AI Agent from the main workflow

Input data:

json

{

“file_id”: “BqacagiaAxkbaambapgkeesdnvjqq_d7rlml8rcs_f0aamkhaai4kmblg9jxvdp4rlm2ba”

}

2. Download File

Downloading a character image from Telegram

• Resource: file
• File ID from the input
• Output: Binary data images

3. Extract & Convert

Preparing an image for AI models

• Conversion to base64
• Preserving the original format
• MIME Type: image/png or image/jpeg

4. Analyze Character

OpenAI Vision creates an ambassador profile

• Model: chatgpt-4o-latest
• Temperature: 0.7

Profile structure:

• Primary identity (name, age, location, profession)
• Appearance and style
• Personality and communication
• Way of life
• The basis of trust

Result: 500+ words of character description

5. Build Video Prompts

Creating a master prompt for video

UGC video aesthetics:

• Freehand camera shake
• Natural focus switching
• Real lighting
• Cluttered background
• Authentic imperfections

Frame-by-frame structure:

• 0-2 sec: Mid-conversation clue
• 2-9 seconds: Product demonstration
• 9-12 sec: Natural completion

6. Generate Ad Script

Gemini 2.5 Pro creates a frame-by-frame scenario

• A detailed description of every second
• Camera position and motion
• Character actions
• Audio replicas

7. Clean Script

GPT-4.1 clears the script from technical comments

• Model: gpt-4.1
• Temperature: 0.3 (minimum creativity)

8. Generate First Frame

Adapting an image to a vertical format

• Gemini Flash expands the background
• Final size: 720x1280
• Save the original composition

9. Sora 2 Video Generation

Main video generation:

json

{

“url”: "https://api.openai.com/v1/videos “,

“model”: “sora-2",

“seconds”: 12,

“size”: “720x1280",

“prompt”: “[cleared script]”,

“input_reference”: “[base64 first frame]”

}

10. Status Monitoring Loop

Readiness monitoring:

• Wait 15 seconds
• Status check
• When completed, download
• When processing, repeat the cycle
• Maximum: 40 iterations (10 minutes)

11. Upload to Google Drive

Save the finished video:

• Title: Video_ [timestamp] .mp4
• Returning WebContentLink for publication

Output data:

json

{

“WebContentLink”: "https://drive.google.com/uc?id =... “,

“WebViewLink”: "https://drive.google.com/file/d/... “,

“duration”: 12,

“fileSize”: “15.2MB”

}

SUB-WORKFLOW: AUTO_POSTING

Publication process

1. When Executed by Another Workflow

Gets a video link from the main workflow

2. Set Accounts

Platform configuration:

json

{

“blotato_api_key”: “Blotato API key”,

“tiktok_id”: “TikTok account ID”,

“video_caption”: “Video description”,

“instagram_id”: “optional”,

“youtube_id”: “optional”,

“facebook_id”: “optional”

}

3. Upload to Blotato

Uploading videos to Blotato servers:

• URL: https://backend.blotato.com/v2/media
• Method: POST
• Returns the internal URL for publishing

4. Publish to TikTok

Final publication:

json

{

“target”: {

“targetType”: “tiktok”,

“PrivacyLevel”: “PUBLIC_TO_EVERYONE”,

“DisabledDuet”: false,

“DisabledComments”: false

},

“content”: {

“text”: “Description and hashtags”,

“MediaURLS”: ["Video URL"]

},

“accountID”: “tiktok_id”

}

```

5. Send Confirmation

Telegram confirmation: “Video published”

Multiplatform support

By changing TargetType, you can publish to:

- Instagram Reels (TargetType: “instagram”)

- YouTube Shorts (TargetType: “youtube”)

- Facebook Reels (TargetType: “facebook”)

- Twitter/X (TargetType: “twitter”)

- LinkedIn (TargetType: “linkedin”)

- Pinterest (TargetType: “pinterest”)

---

USAGE SCENARIOS

Successful generation:

```

You are: [Send a photo]

Bot: “The material is being processed...”

[5-7 minutes of generation]

Bot: “The video is ready! Post to TikTok?”

You: 'Yes'

Bot: “Video posted”

```

Refusal to publish:

```

Bot: “The video is ready! Post to TikTok?”

You: 'No'

Bot: “Good! Video saved: [Google Drive link]”

METRICS AND EFFICIENCY

Time to complete the steps:

• Image loading: 1-2 seconds
• AI character analysis: 15-20 seconds
• Scenario generation: 10-15 seconds
• Frame preparation: 5-8 seconds
• Sora 2 generation: 8-10 min
• Loading to Drive: 3-5 sec
• Publication: 5-10 sec

Common metrics:

• Full time: 5-7 minutes
• Success rate: 94%
• Video price: $2-3
• Platform support: 11+

SYSTEM SETUP

Required services:

1. Telegram Bot
   
   • Create via @BotFather
   • Get a token
   • Configure in N8n
2. OpenAI API
   
   • GPT-4.1-mini access key
   • Access to Sora 2
3. Blotato Account
   
   • Register on blotato.com
   • Obtaining an API key
   • Connecting social media
4. Google Drive
   
   • OAuth2 setup
   • Video folder

ERROR HANDLING

Error types and solutions:

• 400: Wrong image format - check MIME type
• 413: File is too big — 10MB limit
• 429: Rate limit - wait 60 seconds and repeat
• 500: Server error - retry after 30 seconds

Retry logic:

• Maximum 3 attempts
• Increasing the interval between attempts
• Fallback to save to Drive without publishing

RESULT

The system turns Telegram into a full-fledged video studio, where you just need to send a photo to receive a ready-made promotional video posted on TikTok in a few minutes.