Animated logo video from a Google Sheets queue

Animated logo video from a Google Sheets queue

An n8n workflow. List your logos in a Google Sheet, mark one for production and run it. OpenAI reads the logo, an agent
plans a short animation, Veo 3 Fast on Kie.ai renders it, and the clip's address goes back into the row.

What is inside

• workflow.json: the workflow as n8n exports it, 13 nodes. It holds no credential. The spreadsheet and the Kie.ai key
  are left as placeholders.
• README.md: these notes.

Before you start

• An n8n instance, on n8n Cloud or your own server.
• An OpenAI API key.
• A Kie.ai API key.
• A Google account for the sheet, and the logos as image files in Google Drive.
• A spreadsheet with a tab named Sheet1 and these column names in row 1: id, input_logo, status and
  output_logo.

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. Fill the sheet with one row per logo: a unique id, the logo's Google Drive link in input_logo and
   for production in status. Leave output_logo empty.
3. In Google Drive, share each logo so that anyone with the link can view it. OpenAI and Kie.ai fetch the image
   through a download address made from that link.
4. Copy the spreadsheet's ID, the long part of its address between /d/ and /edit. Add a Google Sheets OAuth2
   credential, select it in Get input logo and Store output, and replace YOUR_SHEET_ID with the ID in both.
5. Add an OpenAI credential and select it in Analyze Image and GPT.
6. Open Analyze Image and pick a current vision model, such as gpt-4o. The file names chatgpt-4o-latest, which
   OpenAI's deprecations page lists as removed from the API on 17 February 2026. In the same node, switch off Simplify
   Output. The agent reads the logo description from choices[0].message.content, and only the full output holds it
   (see Good to know).
7. In Create Veo3 and Get Veo3, replace <YOUR_KIE_API_KEY> in the Authorization header with your key. The
   brackets go too.
8. Save and press Execute workflow.

How a run goes

• Your creative direction holds four settings: the brief, the frame (horizontal), the Veo model (veo3_fast) and
  the voice-over wish (no voiceover please).
• Get input logo returns the first row whose status is for production. Code in JavaScript1 takes the file ID
  out of the Drive link.
• Analyze Image sends the logo to the vision model you picked in step 6. It answers in YAML: the brand name, each
  color as a hex code with a name, the font style and a sentence or two on what the logo shows.
• Logo Animator AI Agent runs on gpt-4.1 and has a Think tool to check its answer. Its system message asks for a
  clip of 7.8 seconds in three scenes, each with its camera move, sound and background elements. Structured Output
  holds the answer to three fields: aspect_ratio, video_model and full_video_prompt.
• Create Veo3 posts the plan, the model, the frame and the logo's download address to Kie.ai. Wait1 pauses 250
  seconds, then Get Veo3 asks for the result once.
• Store output finds the row by its id and writes done in status and the clip's address in output_logo.

Change the look

• The brief: your_creative_direction in Your creative direction. As it comes, it asks for an on-brand animation
  with moving background elements and a lively camera.
• Frame: aspect_ratio in the same node. Write vertical for a portrait clip. The agent may only answer 16:9 or 9:16.
• Model: model, veo3_fast or veo3.
• Voice: dialogue. Leave it empty and the agent sets each scene's dialog to null.
• Length and scenes: the agent's system message, which fixes 7.8 seconds and the three-scene frame.
• Render wait: Wait1, 250 seconds.

Good to know

• Each run takes one row. To work through a queue, press Execute workflow once per row, or swap the manual trigger
  for a Schedule Trigger. A run that finds no row marked for production stops after Get input logo.
• Analyze Image keeps n8n's default Simplify Output, which returns only content, while the agent's prompt reads
  choices[0].message.content. Left that way, the agent gets an empty logo description or stops with an error,
  depending on your n8n version. Step 6 fixes it, and so does changing the prompt to $json.content.
• Get Veo3 asks once. If Veo has not finished by then, Store output has no address to write, and the
  row can end up marked done with an empty output_logo. Raise Wait1, set the row back to for production and run
  it again.
• Create Veo3 puts the agent's plan into its body without quotes, so it arrives as a JSON object. Kie.ai's docs,
  read on 2 November 2026, describe the prompt as text. The logo's address also goes in imageUrls alone, not in a
  list. If Kie.ai rejects the request, wrap the plan in JSON.stringify() and the address in square brackets.
• Each run is billed by OpenAI for the image read and the agent's calls, and by Kie.ai for the render.
• Each logo stays shared by link, and it goes to OpenAI and Kie.ai.
• The Kie.ai key sits in two request headers as plain text, so it travels with any export of the workflow. Take it
  out before you share the file.
• The original copy's Code in JavaScript1 code had lost the backslashes of its regex, which broke the node. This one
  has them back: /\/d\/([A-Za-z0-9_-]+)/.