Cinematic VFX video ad from a product photo

What it is

This n8n workflow turns one picture of a product or a logo into a short cinematic ad. Eleven nodes run in a line. A form takes the picture, a business name, a description and an optional video idea. Cloudinary stores the picture at a public address, and OpenAI judges whether it shows a product or a logo and describes it. An agent, briefed as a Hollywood VFX director, writes the video brief as JSON. Two code nodes shape that brief for Kie.ai, which renders it with Veo 3 Fast. After a three-minute wait the workflow asks for the result and downloads the clip.

1. A form takes the picture and the business details
2. Cloudinary stores it and OpenAI describes it
3. A VFX director agent writes the video brief
4. Veo 3 Fast on Kie.ai renders the clip
5. The clip is downloaded after a three-minute wait

Who it suits

Small brands and shops that want a dramatic motion ad from a single picture, and the marketers who make those ads for them. The director’s brief leans toward food and drink, with ingredients assembling in slow motion. Other products work, but you may want to rewrite the brief.

Skip it if you want a calm, true-to-life product clip. The brief asks for particles, volumetric light and bold camera moves, which makes a loud result.

Where to use it

Any n8n that can reach Cloudinary, OpenAI and Kie.ai.

• n8n Cloud: import the file into a new workflow there
• Your own n8n server: the same file, the same steps
• Ad and social channels: the clip comes back as a file, ready for an ad, a Reel or a Short
• A drive or another app: add a node after HTTP Request2 to save the clip or send it on

Setup

Get it saves a zip with workflow.json and Mindket’s setup notes in your downloads folder. The file holds no credential, and it calls no sub-workflow.

1. In n8n, create a workflow, open its menu, choose Import from File and pick workflow.json.
2. Open Upload Photo1. In its address, put your cloud name where it says <YOUR_CLOUD_NAME>. Swap YOUR_PRESET_NAME for your preset and YOUR_FOLDER_NAME for a folder.
3. Add an OpenAI credential and select it in Analyze image and OpenAI Chat Model.
4. In HTTP Request and HTTP Request1, replace <YOUR_KIE_API_KEY> after Bearer with your Kie.ai key.
5. Save, send a test picture through the form’s test address, then press Publish.

How to use it

Open the form from VFX Ad Video, attach a product photo or a logo and fill in the fields. The form confirms at once. The run goes on behind it: upload, analysis, brief, render, download. The clip lands in the run’s execution data as a file, so open the run under Executions to see it.

To change the look, edit the system message in AI Agent. The extra product wording sits in Code in JavaScript1, the frame in aspectRatio in HTTP Request, and the render wait in Wait.

Example

A photo of a pack of oat biscuits, with the description “small-batch oat biscuits” and no video idea. OpenAI is asked to label the picture a product or a logo and to describe its colors, packaging and mood. The agent answers with a JSON brief of this shape:

{
  "description": "…",
  "style": "…",
  "camera": "…",
  "lighting": "…",
  "environment": "…",
  "elements": ["…"],
  "motion": "…",
  "ending": "…",
  "text": "none",
  "keywords": ["16:9", "…"]
}

The description and the camera line are joined, with extra effects wording added when the analysis reads as a product, and sent to Kie.ai with the photo’s address.

For newcomers and experts

If you are new. n8n draws a workflow as boxes joined by lines, and each box, a node, does one job. A credential is a key n8n keeps for you and hands to the node you pick it in. Run the form from its test address first. Every node then shows what it received and what it sent, which is where a wrong preset or key turns up.

For experts. One OpenAI model does both AI jobs, and Kie.ai does the render.

The product test in Code in JavaScript1 is a keyword search of the analysis for words such as product, food, packaging, ingredients, cookie and snack. A logo whose analysis mentions a product gets product effects. The first code node’s output is not used by the second, which reads the agent and the analysis itself.

Good to know

Each run costs money, and the render time is a guess.

• Billing: OpenAI and Kie.ai bill their own part of a run
• One wait: if the render is not ready after the three minutes, there is no address to download and the run fails, so raise the wait for long clips
• Open form: once the workflow is published, anyone with the form’s address can start a run on your keys, so turn on the form’s basic authentication or keep the address to yourself
• The picture: an unsigned upload makes it public on Cloudinary, and it also goes to OpenAI and Kie.ai
• Quotes: JSON.stringify wraps the prompt and the watermark in quotation marks, so the marks may show up in the text Kie.ai receives
• The model name: chatgpt-4o-latest is a moving alias that OpenAI may retire, so pick another vision model in both nodes if one says it is gone
• File types: the form’s file field takes any file type, so send pictures only