Auto-Generate UGC Video Ads from Product Images with n8n, Kie.ai Sora 2 & Tavily

Enter product name, prompt, and image link in Google Sheets → AI Agent with Tavily research crafts a video prompt → Kie.ai Sora 2 outputs a 10-second portrait ad → writes result back to Sheets

Who

E-commerce operators, UGC marketers, AI automation freelancers, and product teams looking to batch-generate short video ads

You get

After filling product info in Google Sheets, trigger the n8n workflow to auto-research trends, generate an optimized prompt, call Kie.ai Sora 2 for a 10-second watermark-free portrait video, and write the link back to the sheet

You need

n8n · google-sheets · tavily · kie.ai

Steps

1. Prepare Google Sheets and product data

   Create or open a Google Sheets with columns: ID, Product Name, Prompt, Image Link, Video Link, Processed. Fill in product names, short prompts or ideas, and image links, leaving Processed blank or set to No so the workflow can identify unprocessed rows.

2. Configure credentials and Header Auth

   In n8n, create and connect the following credentials: - Google Sheets OAuth 2.0 - OpenAI API Key - Tavily Search API: use Generic Credential Type → Header Auth, Name = Authorization, Value = Bearer YOUR_TAVILY_TOKEN - Kie.ai Sora 2 API: same Header Auth setup, Name = Authorization, Value = Bearer YOUR_KIEAI_TOKEN Bind each credential to its corresponding node.

3. Import the workflow and link your sheet

   Download the template workflow JSON and import it into n8n. Open the 'Get row(s) in sheet' and 'Append or update row in sheet' nodes, replacing the documentId and sheetName with your own Google Sheets ID and sheet name. Verify that column mappings match your sheet.

4. Execute workflow: read an unprocessed product row

   Click 'Execute Workflow' to trigger manually. The workflow searches Google Sheets for the first row where Processed is empty/No, reads Product Name, Prompt, Image Link, and passes them to downstream nodes.

5. AI Agent generates video prompt (with Tavily research)

   The AI Agent node receives product info, first calls Tavily to search for latest trends and marketing insights for the product category, then uses the OpenAI Chat Model (gpt-4.1-mini) to generate a detailed Sora 2 video prompt covering subject appearance, setting, camera style, tone, and lighting.

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

   Copy prompt

   =You are an expert AI video prompt engineer trained to design optimized prompts for Sora 2 video generation on Kie.ai.
   Your role is to take a raw product concept, a short idea, and an image link and transform them into a highly detailed, realistic, and visually rich video prompt that aligns with modern marketing standards.
   
   
   🎯 Objective
   Generate a prompt that helps Sora 2 produce a realistic, natural-looking short video matching the intent of the product or service.
   The result should look like it was filmed by a real person or a professional ad team — depending on the desired style.
   
   
   🧩 Your Rules
   1. Before generating the video prompt, always use the Tavily tool to search for the latest trends, product type, related products or categories, and current market insights of the product or product category. Use this information to ensure the video prompt reflects the most current and effective styles in video advertising, avoiding outdated approaches.
   2. Understand the product — analyze what it is, who it’s for, and its emotional appeal (use any research insight provided).
   3. Always describe:
   - Main subject(s): appearance, clothing, expression, action.
   - Setting: location, lighting, time of day, atmosphere.
   - Camera style: framing, motion, handheld vs cinematic, lens type, realism.
   - Tone: emotional mood (friendly, elegant, cinematic, testimonial, etc.).
   - Lighting: realistic natural or studio lighting, with mood consistency.
   4. If the user specifies “UGC” or “influencer-style”, simulate a smartphone or handheld camera perspective, vertical 9:16 format, and a natural, authentic tone.
   5. If no style is specified, default to a cinematic ad style that looks visually stunning and brand-ready.
   6. Avoid repetition, keep every sentence meaningful, and ensure the description reads like a cinematographer describing the shot to a director.

6. Call Kie.ai Sora 2 to generate video

   The 'Create Video' node sends the AI-generated prompt and product image to the Kie.ai API (https://api.kie.ai/api/v1/jobs/createTask). Request parameters include: model = sora-2-image-to-video, aspect_ratio = portrait, n_frames = 10, remove_watermark = true. If the image link uses a Google Drive share format, the node automatically extracts the file ID and converts it to a direct download link.

7. Wait and poll for video result

   The workflow pauses for 15 seconds via the Wait node, then the 'Grab Video' node queries Kie.ai task status using the taskId. The Switch node evaluates the response: - success → extract video link and continue - 500 error → log error message - still pending → loop back to Wait node

8. Write back to Google Sheets and mark complete

   On success, the 'Edit Fields' node extracts the watermark-free video link from resultJson, and the 'Append or update row in sheet' node writes Video Link and Processed = Yes back to the original row, completing the cycle.

You are setting up an n8n workflow to automatically generate UGC-style short video ads from product data in Google Sheets. First confirm you have: an n8n instance, a Google Sheets with the required columns, Google Sheets OAuth 2.0 credential, OpenAI API Key, Tavily API Key (Header Auth), and Kie.ai Sora 2 API Key (Header Auth). Then import the workflow, replace the sheet ID, bind credentials, and manually execute to test one row.