Studio, lifestyle, close-up and action shots of one product

Studio, lifestyle, close-up and action shots of one product

An n8n workflow that turns one product photo into four new ones for a shop listing. You upload the photo through a form and
choose an aspect ratio. The photo goes to Cloudinary, and Gemini describes it. Gemini then writes four image prompts:
a studio shot on a clean background, a lifestyle shot in use, a close-up of the details and an action shot from a dynamic
angle. Kie.ai's Nano Banana Edit model makes each image from your photo and its prompt, and the workflow fetches the four
finished files.

What is inside

• workflow.json: the workflow as n8n exports it, twenty-nine nodes with no sticky notes. It holds no credential.
• README.md: these notes.

What we changed from the original copy

• API Values had three empty fields and a note asking for a key. They now read YOUR_CLOUD_NAME, YOUR_UPLOAD_PRESET and
  YOUR_KIE_API_KEY, so each value you type has a clear place. folder stays empty, since it is optional.
• Nothing else changed. The original copy needed no backslash repair.

Before you start

• An n8n instance, on n8n Cloud or your own server.
• A Gemini API key from Google AI Studio, with access to gemini-2.5-pro (see Good to know).
• A Cloudinary account with an unsigned upload preset.
• A Kie.ai account, its API key and credits.

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. In the Cloudinary console, copy your cloud name from the dashboard. Under Settings, Upload, add an upload preset with
   Signing Mode set to Unsigned, and note its name.
3. Open API Values. Replace YOUR_CLOUD_NAME and YOUR_UPLOAD_PRESET with those two values, and YOUR_KIE_API_KEY with
   your Kie.ai API key. Type a folder name in folder if you want the uploads kept in one.
4. Add a Google Gemini (PaLM) API credential with your Gemini key and select it in Analyze an image and
   Google Gemini Chat Model.
5. Save, open On form submission and send a photo through its test address. Watch the four branches finish.
6. Press Publish to put the form at its live address.

How a run goes

1. On form submission takes one product image and an aspect ratio: 9:16, 16:9 or 1:1.
2. getBinaryFile keeps the photo, API Values adds your account values, and Merge joins the two.
3. Analyze Img uploads the photo to Cloudinary with your unsigned preset and gets back its web address.
4. Analyze an image asks Gemini what is in the photo. The node's own default question is used, as the file sets
   none.
5. Basic LLM Chain hands that description to Gemini with a long brief and gets four prompts back as JSON:
   1stPrompt for the studio shot, 2ndPrompt for lifestyle, 3rdPrompt for the close-up and 4thPrompt for the action
   shot. It retries if the answer fails.
6. Four branches then run side by side. Create Image to Create Image3 each start a Kie.ai task with one prompt, your
   photo's address and the aspect ratio, asking for a PNG.
7. Each branch waits 30 seconds, asks Kie.ai for the task's state, and waits again until the state is success.
8. Load Image File to Load Image File3 download the finished images.

Change the look

• The four styles: the brief in Basic LLM Chain. Its four numbered shot types and the detail list under them shape every
  prompt.
• The description: add your own question in Analyze an image, such as the product's name and material, to steer the prompts.
• The sizes: the options of Aspect Ratio in the form. Kie.ai also takes 3:4, 4:3, 4:5, 5:4, 2:3, 3:2 and 21:9.
• The file type: output_format in the four Create Image bodies, png or jpeg.
• How often it checks: the four Wait nodes.

Good to know

• Where the photos end up: nowhere yet. The four files sit in the Load Image File nodes of the run, so open it in Executions
  to download them. To keep them, add a Google Drive or Cloudinary upload after each of those nodes. Kie.ai keeps its copies
  for 14 days.
• A failed image never ends: each If only lets success through, so a task that ends in fail keeps the branch polling
  every 30 seconds. Kie.ai also answers a task it refuses with status 200, an error code and no task ID, and that branch
  loops the same way. Set a timeout in the workflow's settings, or stop such a run by hand.
• Quotes in prompts: each prompt goes into the request body between quotation marks. A prompt with a double quote or a line
  break breaks that request, and the run stops with an error. Writing {{ JSON.stringify($json.output['1stPrompt']) }} in
  place of the quoted expression, and the same for the other three, avoids it.
• image_size: Kie.ai's page for the model, read on 3 November 2026, marks it as replaced by aspect_ratio. If every photo
  comes back square, rename the field to aspect_ratio in the four Create Image bodies.
• The Gemini model: both Gemini nodes use gemini-2.5-pro. Google's model notes, read on 2 November 2026, gives it no shutdown
  date, but limits that model family to projects that already used it. If your key gets an access error, pick a current model,
  such as gemini-3.8-flash, in both nodes.
• Billing: Gemini bills two calls a run, the analysis and the prompts, and Kie.ai takes credits for each of the four images.
  Cloudinary keeps every uploaded original, which counts against your plan.
• Keys in plain sight: the Kie.ai key sits in API Values, so anyone who can open the workflow can read it.
• Who can use it: once published, anyone with the form's address can spend your Kie.ai credits. An unsigned preset also lets
  anyone who knows your cloud name and preset name upload to your Cloudinary.
• Your data: the photo is stored in Cloudinary at a public address and sent to Google and Kie.ai. Check that the generated
  photos match the product you sell before you list them.