Auto-Generate E-Commerce Promotional Carousels with n8n (Gemini 2.5 + Social Publishing)

Upload product image & description → AI generates 4 promo images (prompts target vertical 9:16) → auto-store & optionally publish to Instagram/TikTok/YouTube/Facebook

Who

E-commerce operators, digital marketers, social media managers

You get

Generate 4 promo images with real-people usage scenes (prompts target vertical 9:16; the request sets no dimensions; not tested by us) from a single product photo, plus social copy, with optional auto-publishing

You need

n8n · fal.ai · imgbb · upload-post.com

Steps

1. Upload product image and description via form

   Before the first test, disable or disconnect the 'Upload Post' node (see step 5) so a test run cannot publish to social media. Then trigger the 'Photo Upload Form' node (formTrigger) in n8n. Upload a product image and fill in the 'Product description' text area. On submission, the workflow splits data into two branches: one enters AI storyboard analysis, the other merges image variables.

2. AI generates 4-image storyboard (Storyboard Agent)

   The 'Storyboard Agent' node calls Google Gemini 2.5 Pro (via googlePalmApi credential) to analyze the product description and output a strict JSON storyboard containing: title (≤60 chars), 4 image generation prompts (prompt1-prompt4), environment description, and sound description. A Structured Output Parser enforces JSON compliance; a Think tool assists reasoning.

   Execution prompt · Source text。Storyboard Agent node System Message field

   Copy prompt

   Role: You are a creative director specialized in promotional ads for social media. You will produce a 4-image carousel (vertical 9:16) to promote products. The goal is to create authentic content that generates engagement and conversions.
   
   Content rules:
   
   - ALWAYS include people using the product naturally
   - Add visible promotional text in images (offers, benefits, CTAs)
   - Show product in real usage situations
   - Create attractive compositions for Instagram/TikTok/Facebook
   - Maintain visual coherence between the 4 images
   - Focus on generating engagement and conversion
   - Output strictly valid compact JSON, no prose, no markdown.
   
   Language: Follow the input description language (ES/EN).

   Execution prompt · Source text。Storyboard Agent node Text/Prompt field

   Copy prompt

   ==Use this product description:
   
   {{ $json["Product description"] }}
   
   Create a promotional carousel of 4 images for social media. Focus on showing the product IN USE by real people with promotional text.
   
   Your task:
   1. Design 4 independent images showing different aspects of the product
   2. Include people using/enjoying the product naturally
   3. Add promotional text elements, benefits or call-to-actions
   4. Create an engaging carousel for Instagram/TikTok/Facebook
   
   For the product context, consider:
   - Real people using the product in everyday situations
   - Visible promotional text (offers, benefits, CTAs)
   - Authentic environments related to product usage
   - Compositions that generate engagement and conversion
   
   Return ONLY a compact JSON with these keys (no prose):
   {
     "title": "<≤60 chars, catchy hook title that captures attention>",
     "prompt1": "<Image 1: Person using the product - vertical 9:16, includes visible promotional text, natural environment>",
     "prompt2": "<Image 2: Key product benefit in use - shows results or satisfaction, with benefit text>",
     "prompt3": "<Image 3: Product in lifestyle context - person enjoying, text with offer or CTA>",
     "prompt4": "<Image 4: Hero shot of product with person - impactful final composition, text with call-to-action>",
     "i2v_prompt1": "Static promotional image for carousel - no animation needed",
     "i2v_prompt2": "Static promotional image for carousel - no animation needed", 
     "i2v_prompt3": "Static promotional image for carousel - no animation needed",
     "i2v_prompt4": "Static promotional image for carousel - no animation needed",
     "environment": "<≤25 words: authentic environment where product is used, real situations>",
     "sound": "<≤20 words: modern catchy commercial music - suitable for social media ads>"
   }
   
   Rules:
   • INCLUDE people using the product naturally and authentically
   • ADD visible promotional text in images (offers, benefits, CTAs)
   • Show product in real usage situations
   • Focus on generating engagement and conversion
   • Attractive compositions for social media
   • Maintain visual coherence between the 4 images
   • Language should follow input description language
   • Output strictly valid JSON

3. Generate 4 promo images in parallel and upload to imgbb

   After 'Set Storyboard Vars' decomposes the JSON into variables, the workflow executes 4 parallel branches. Each branch calls fal.run/fal-ai/gemini-25-flash-image/edit via HTTP Request (POST), passing the corresponding prompt and original product image URL to generate 1 image. Then Split Out separates results, a Code node renames binary data to photo2-photo5, and each is uploaded to imgbb for a public URL. The 4 branches merge into an array.

   Execution prompt · Source text。Gemini 2.5 Flash - Generate Image 2/3/4/5 node Body field (raw JSON); prompt1-prompt4 map to 4 nodes respectively

   Copy prompt

   ={"prompt":"{{ $('Set Storyboard Vars').item.json.prompt1 }}","image_urls":["{{ $('Upload Original Image to imgbb').item.json.data.url }}"],"num_images":{{ $('Set APIs Vars').item.json.number_of_images }}}

   Execution prompt · Source text。Rename to photo 2 node JS Code (photo3/4/5 analogous)

   Copy prompt

   return items.map((i)=>({json:{url: i.json.url},binary:{photo2:i.binary.data}}));

4. Generate social media promotional copy

   The 4 imgbb URLs are aggregated via Merge + Aggregate, then fed into the 'Generate Carousel Description' OpenAI node. Using GPT-4.1, it generates an Instagram/TikTok promo description (≤150 chars) based on the 4 prompts, matching the input description language.

   Execution prompt · Source text。Generate Carousel Description node System Message field

   Copy prompt

   You are an expert at creating descriptions for Instagram/TikTok carousels.

   Execution prompt · Source text。Generate Carousel Description node User Message field

   Copy prompt

   =I'm going to upload a carousel of 4 promotional images of a PRODUCT (9:16). Create an advertising description for Instagram/TikTok based on these prompts:
   1) {{ $('Set Storyboard Vars').item.json.prompt1 }}
   2) {{ $('Set Storyboard Vars').item.json.prompt2 }}
   3) {{ $('Set Storyboard Vars').item.json.prompt3 }}
   4) {{ $('Set Storyboard Vars').item.json.prompt4 }}
   
   IMPORTANT: Respond only with the description (≤ 150 characters), benefit-oriented with a strong CTA. If the prompts are in Spanish, respond in Spanish; if in English, respond in English.

5. Optional: auto-publish to social platforms

   If the upload-post.com credential is configured, the 'Upload Post' node combines the 4 imgbb image URLs with AI-generated copy and auto-publishes to Instagram, TikTok, YouTube, and Facebook. You must set your upload-post.com username, Facebook Page ID, and other parameters in the node. Note: the source template already wires this node directly after the copy node, so it is not off by default and a manual run can publish; disable or disconnect it before the first test and reconnect only after checking the images and copy.

This is an official n8n workflow template. Before building, ensure you have: fal.ai API token, imgbb API key, OpenAI API key; optionally upload-post.com account and Instagram business account. After importing the template into n8n, configure credentials and variables per the steps above. Test with a single product image and short description first to validate storyboard output and image generation.