Thursday, December 11, 2025 / Clicky News
Google Vertex AI : Automating Creative Excellence
Google Vertex AI : Automating Creative Excellence
Nano Banana Pro and JSON prompts are one thing. How do we deliver this kind of studio quality image production at scale I hear you ask? Well lets talk about Vertex AI from Google.
Nano Banana Pro and JSON prompts are one thing. How do we deliver this kind of studio quality image production at scale I hear you ask? Well lets talk about Vertex AI from Google.




Nathan Smith
Nathan Smith
Head of AI
Head of AI
In our previous post, we explored how to achieve studio-quality imagery using Nano Banana Pro and structured JSON prompts. We demonstrated that by treating creative direction as data – defining specific key-value pairs for lighting, subject, and product – we can generate assets that are virtually indistinguishable from a high-end editorial shoot.
Below is an image of the photoshoot that never happened…

But creating one perfect image is just the beginning. The real challenge for modern brands is scale. How do you take that single, perfectly crafted aesthetic and roll it out across thousands of SKUs, multiple territories, and infinite user segments without blowing the budget?

This is where we move from the studio to the engine room. This is where Google Vertex AI changes the game.
From Creation to Proliferation with Google Vertex
If Nano Banana Pro is the photographer, Google Vertex AI is the global production house that never sleeps. While our first part focused on the fidelity of the image, Part 2 is about the utility of the workflow.

Vertex AI allows us to take the logic we established with our JSON prompts and automate it. Instead of manually tweaking the code for "Sage Green" or "Blush Pink", we can connect the generative model directly to your product feed.
Here is how we utilise this infrastructure to proliferate content for a brand:
1. Dynamic Prompt Injection We no longer write individual prompts. Instead, we build a prompt architecture within Vertex AI. We map the fields in your product catalogue (Colour, Material, SKU, Season) to the JSON structure we defined in Part 1.
When a new product line drops, the system automatically populates the product and subject fields in the JSON. Vertex AI then instructs the model to generate the corresponding imagery for every single item, ensuring the lighting ("Soft studio lighting, 50mm f/2.8") remains consistent across the entire range.
2. Contextual Variation Proliferation is not just about cloning; it is about relevance. We can configure Vertex to generate variations of the scene based on audience data.
For the "Wellness" audience: The system injects a "calm, morning yoga studio" background into the
scenevariable.For the "High Performance" audience: The system swaps this for a "dynamic, high-contrast gym environment".
The product remains identical. The brand guidelines are strictly adhered to. But the context changes instantly to match the user's intent.
3. Format Adaptation In a traditional workflow, resizing a campaign image for a 9:16 Story, a 1:1 Feed post, and a wide website banner often involves awkward cropping or compromise.
With Vertex AI, we don't crop. We regenerate. The system understands the compositional needs of each placement. It extends the background seamless paper, adjusts the depth of field, or moves the subject to ensure the asset is native to the platform it will live on. We are generating bespoke creative for every placement, not just resizing a master file.
Safety & Brand Integrity
One of the biggest fears regarding automated content is the "hallucination" risk – the AI generating something odd or off-brand.
This is why we use Vertex AI’s robust safety filters and "grounding" capabilities. We can enforce negative prompts (e.g., "no distorted text", "no competitors") at an architectural level. Furthermore, we can implement a "Human-in-the-Loop" stage where a creative director reviews a sample set before the system batches out the remaining 5,000 assets. You get the speed of AI with the assurance of human oversight.
The New Creative Efficiency
The goal here is not to replace the creative team but to free them from the repetitive drudgery of versioning. By combining the precision of Nano Banana Pro’s JSON prompting with the industrial scale of Google Vertex AI, we are building a content engine that creates consistent, high-quality, and hyper-relevant imagery at a pace that was previously impossible.

We are moving from "making an ad" to "programming a brand universe".
This kind of technology is not really needed or necessary for smaller simpler brands but for large brands with hundreds of thousands of products and hundreds of territories and campaigns the temptation to utilise this tech will be a no-brainer.
Are you ready to scale your creative output? If you want to learn more about automating your brand’s content production using Vertex AI, get in touch with our team today.
In our previous post, we explored how to achieve studio-quality imagery using Nano Banana Pro and structured JSON prompts. We demonstrated that by treating creative direction as data – defining specific key-value pairs for lighting, subject, and product – we can generate assets that are virtually indistinguishable from a high-end editorial shoot.
Below is an image of the photoshoot that never happened…

But creating one perfect image is just the beginning. The real challenge for modern brands is scale. How do you take that single, perfectly crafted aesthetic and roll it out across thousands of SKUs, multiple territories, and infinite user segments without blowing the budget?

This is where we move from the studio to the engine room. This is where Google Vertex AI changes the game.
From Creation to Proliferation with Google Vertex
If Nano Banana Pro is the photographer, Google Vertex AI is the global production house that never sleeps. While our first part focused on the fidelity of the image, Part 2 is about the utility of the workflow.

Vertex AI allows us to take the logic we established with our JSON prompts and automate it. Instead of manually tweaking the code for "Sage Green" or "Blush Pink", we can connect the generative model directly to your product feed.
Here is how we utilise this infrastructure to proliferate content for a brand:
1. Dynamic Prompt Injection We no longer write individual prompts. Instead, we build a prompt architecture within Vertex AI. We map the fields in your product catalogue (Colour, Material, SKU, Season) to the JSON structure we defined in Part 1.
When a new product line drops, the system automatically populates the product and subject fields in the JSON. Vertex AI then instructs the model to generate the corresponding imagery for every single item, ensuring the lighting ("Soft studio lighting, 50mm f/2.8") remains consistent across the entire range.
2. Contextual Variation Proliferation is not just about cloning; it is about relevance. We can configure Vertex to generate variations of the scene based on audience data.
For the "Wellness" audience: The system injects a "calm, morning yoga studio" background into the
scenevariable.For the "High Performance" audience: The system swaps this for a "dynamic, high-contrast gym environment".
The product remains identical. The brand guidelines are strictly adhered to. But the context changes instantly to match the user's intent.
3. Format Adaptation In a traditional workflow, resizing a campaign image for a 9:16 Story, a 1:1 Feed post, and a wide website banner often involves awkward cropping or compromise.
With Vertex AI, we don't crop. We regenerate. The system understands the compositional needs of each placement. It extends the background seamless paper, adjusts the depth of field, or moves the subject to ensure the asset is native to the platform it will live on. We are generating bespoke creative for every placement, not just resizing a master file.
Safety & Brand Integrity
One of the biggest fears regarding automated content is the "hallucination" risk – the AI generating something odd or off-brand.
This is why we use Vertex AI’s robust safety filters and "grounding" capabilities. We can enforce negative prompts (e.g., "no distorted text", "no competitors") at an architectural level. Furthermore, we can implement a "Human-in-the-Loop" stage where a creative director reviews a sample set before the system batches out the remaining 5,000 assets. You get the speed of AI with the assurance of human oversight.
The New Creative Efficiency
The goal here is not to replace the creative team but to free them from the repetitive drudgery of versioning. By combining the precision of Nano Banana Pro’s JSON prompting with the industrial scale of Google Vertex AI, we are building a content engine that creates consistent, high-quality, and hyper-relevant imagery at a pace that was previously impossible.

We are moving from "making an ad" to "programming a brand universe".
This kind of technology is not really needed or necessary for smaller simpler brands but for large brands with hundreds of thousands of products and hundreds of territories and campaigns the temptation to utilise this tech will be a no-brainer.
Are you ready to scale your creative output? If you want to learn more about automating your brand’s content production using Vertex AI, get in touch with our team today.
more BLogs
more BLogs





