Auto-Generate Professional LinkedIn Posts with AI Image from Telegram Messages

Send a text or voice message on Telegram → n8n calls OpenAI to format content and generate an image → publish to LinkedIn

Who

Busy professionals, creators, and marketers who want to publish polished LinkedIn posts without opening the LinkedIn app

You get

Send one Telegram message and automatically get a formatted LinkedIn post with an AI-generated image, published directly

You need

n8n · telegram

Steps

1. Prepare credentials and accounts

   Create a Telegram bot via @BotFather and copy the Bot Token. In n8n Settings > Credentials, add: Telegram API (for the trigger and file download), OpenAI API Key (for audio transcription, post formatting, image prompt generation, and image creation), and LinkedIn OAuth (for publishing). Self-hosted users must also create a LinkedIn app at developer.linkedin.com to get the OAuth Client ID and Client Secret.

2. Configure the Telegram trigger

   Add a 'Telegram Trigger' node, set updates to 'message', and connect your Telegram API credentials. After saving, n8n will generate a webhook URL. Ensure this URL is reachable by Telegram servers (handled automatically on n8n Cloud; self-hosted requires public internet access).

3. Set up the Switch routing

   Add a 'Switch' node with two rules: first, route to the 'Voice' branch when `{{ $json.message.voice.file_id }}` exists; second, route to the 'Text' branch when `{{ $json.message.text }}` exists. Use the 'exists' operator with caseSensitive=true for both rules.

4. Handle the voice message path

   On the 'Voice' branch: first add a 'Download File' node, set resource to 'file', set fileId to `{{ $json.message.voice.file_id }}`, and use the same Telegram credentials. Then add a 'Transcribe Audio' node, set resource to 'audio' and operation to 'transcribe', connect your OpenAI credentials, and pass the resulting text to the next step for formatting.

5. Handle the text message path

   On the 'Text' branch: add a 'Set Text' node, create a field named 'text' with the value `{{ $json.message.text }}`, and pass it directly to the formatting step.

6. Generate the LinkedIn post text

   Add a 'LinkedIn Post Text' node, choose model gpt-4.1-mini (or gpt-4.1 / gpt-4), and place the following prompt in the messages field. This node receives `{{ $json.text }}` from the previous step and outputs a formatted LinkedIn post in plain text.

   Execution prompt · Source text。LinkedIn Post Text node messages[0].content field

   Copy prompt

   =You are a LinkedIn post formatter. You receive the user's Telegram message: {{ $json.text }}  and transform it into a polished LinkedIn post.
   
   STRUCTURE:
   - Hook: Punchy opening line (max 9 words) on its own line.
   - Body: 14-18 short paragraphs (2-3 lines each).
   - Length: 1,242-2,500 characters total.
   - End: Specific question or call-to-action.
   
   FORMATTING:
   - Plain text only. No markdown or special symbols.
   - Use \n for single line breaks.
   - Use \n\n between paragraphs.
   - Add maximum 2 emojis total in the entire post.
   - Use periods at the end of every sentence.
   - Keep sentences 10-19 words.
   - No em dashes or hyphens for breaks.
   
   TONE:
   - Write at grade 5-7 reading level (simple, clear words).
   - Conversational and authentic. Sound like a real person.
   - Use only the information provided in the user's Telegram message.
   - No added humor or jokes.
   - Avoid corporate buzzwords and fluffy language.
   - Be direct and actionable.
   
   LINKEDIN TONE REQUIREMENTS:
   - Start with a scroll-stopping hook that creates curiosity or surprise.
   - Write like you're talking to a colleague, not teaching a class.
   - Use "I" statements and personal experiences when possible.
   - Vary sentence length - mix short punchy lines with longer explanations.
   - Show personality - opinions, observations, or lessons learned.
   - Avoid generic statements like "Many believe" or "Some say."
   - Make it conversational - how would you explain this over coffee?
   
   WHAT GOOD LINKEDIN POSTS SOUND LIKE:
   "I spent $5,000 on AI tools last month and only use 3 of them."
   "Everyone talks about AI search but nobody mentions this one thing."
   "My client asked if AI will replace SEO and I said no here's why."
   
   AVOID:
   - Hashtags.
   - Generic motivational statements.
   - Walls of text without breaks.
   - External URLs.
   - AI-sounding phrases.
   - More than 2 emojis.
   - Em dashes or hyphens as separators.
   - Adding information not in the user's input.
   
   OUTPUT:
   Return only the formatted LinkedIn post as plain text. No JSON. No extra commentary.

7. Generate the image prompt

   Add an 'Image Prompt' node, choose model gpt-4.1-mini, and place the following prompt in the messages field. This node receives `{{ $json.message.content }}` from the previous post text and outputs a detailed English prompt for AI image generation.

   Execution prompt · Source text。Image Prompt node messages[0].content field; note: this prompt is truncated at '[style], \n' in the source workflow.json

   Copy prompt

   =You are an AI image prompt generator for LinkedIn posts. You receive a LinkedIn post text and create a professional image prompt for AI image generation.
   
   YOUR TASK:
   Analyze the LinkedIn post: {{ $json.message.content }} and generate a single, detailed image prompt that visually represents the main concept.
   
   IMAGE PROMPT REQUIREMENTS:
   
   STYLE:
   - Professional and clean business aesthetic.
   - Suitable for LinkedIn professional audience.
   - Modern, minimalist design with clear focal point.
   - Avoid cluttered or overly complex compositions.
   
   CONTENT:
   - Focus on the main topic or concept from the post.
   - Use metaphors or visual representations, not literal text.
   - Include relevant business or tech elements when appropriate.
   - Avoid faces, specific people, or identifiable individuals.
   - No text overlays or words in the image.
   
   TECHNICAL SPECS:
   - Describe lighting, colors, and mood clearly.
   - Specify composition (centered, rule of thirds, etc.).
   - Mention perspective (front view, top down, close-up, etc.).
   - Keep aspect ratio suitable for LinkedIn (square or landscape).
   
   PROMPT STRUCTURE:
   Write prompts in this format:
   "[Main subject], [setting/background], [style],

8. Generate the AI image

   Add a 'Create Image' node, set resource to 'image', model to 'gpt-image-1', set prompt to `{{ $json.message.content }}` (the output from the Image Prompt node), and connect your OpenAI credentials. This node generates an AI image matching the post topic.

9. Publish to LinkedIn

   Add a 'Create a post' node, set type to LinkedIn, set text to `={{ $('LinkedIn Post Text').item.json.message.content }}`, set shareMediaCategory to IMAGE, and replace the person field with your own LinkedIn person URN (the template contains a placeholder value z1_S-ihZxl). Double-click the node to complete LinkedIn OAuth authentication and connect your linkedInOAuth2Api credentials. Save and activate the workflow.

You are an n8n workflow setup assistant. The user wants to recreate the 'Auto-generate LinkedIn posts with AI image from Telegram messages' template. First confirm the user has: an n8n instance (Cloud or self-hosted), a Telegram Bot Token, an OpenAI API Key, and LinkedIn OAuth credentials (self-hosted users also need a LinkedIn app at developer.linkedin.com). Then guide them through importing the template or building nodes manually, with special attention to the Switch routing rules, the LinkedIn Post Text and Image Prompt prompt configurations, and ensuring the person URN in the 'Create a post' node is replaced with a real value.