Saturday, December 13, 2025 / Clicky News

Real image to JSON code & edit using Google Nano Banana Pro

We've been busy experimenting with JSON and Google Nano Banana Pro to make images with unbeliveable detail. We now look at how you can deconstruct a real image into JSON and then process it into alternatives.

Oli Yeates

CEO & Founder

We have been busy experimenting with JSON and Google Nano Banana Pro to create images with unbelievable detail. In this post, we look at how you can deconstruct a real image into JSON and then process it into limitless alternatives.

The world of generative AI is moving at a breakneck pace. Just when we think we have mastered standard text prompts, a new method arrives that completely changes the workflow. Recently, we have been exploring a technique that bridges the gap between photography and code: converting real images into structured JSON data to drive the new Google Nano Banana Pro model.

The Challenge of Consistency

One of the biggest hurdles in AI imagery is consistency. If you ask an AI for "a family on a sofa," you get a different result every time. For digital marketers and brands, this randomness can be a problem. We need control. We need to know that if we change the wall colour, the family members won't suddenly turn into different people.

This is where JSON (JavaScript Object Notation) comes in. By converting visual information into code, we create a structured "recipe" for the image that we can edit with precision.

So here is the real image from a stock photography site:

Lets ask Gemini to turn this image into a comprehensive JSON prompt by uploading it and asking "deconstruct this image into a comprehensive JSON prompt".

Here is the output:

{
  "image_description": {
    "summary": "A candid, medium shot of a happy family of three sitting on a teal sofa in a cozy living room, eating popcorn and laughing while watching a movie on a laptop in the foreground.",
    "mood": "Joyful, cozy, domestic, heartwarming, intimate, candid."
  },
  "subjects": [
    {
      "id": "father",
      "demographics": "Adult male, Caucasian, mid-30s",
      "appearance": {
        "hair": "Short, dark brown, styled back",
        "facial_hair": "Light stubble/beard",
        "accessories": "Black-rimmed eyeglasses",
        "expression": "Broad, genuine laugh"
      },
      "clothing": {
        "top": "Denim blue button-down shirt",
        "bottom": "Jeans (partially visible)"
      },
      "pose": "Leaning slightly forward, sitting on the right side of the couch, looking at the screen."
    },
    {
      "id": "mother",
      "demographics": "Adult female, Caucasian, mid-30s",
      "appearance": {
        "hair": "Dark brown, curly, shoulder-length",
        "expression": "Laughing with eyes closed, joyful"
      },
      "clothing": {
        "top": "Mustard yellow knit sweater with pearl details",
        "bottom": "Beige trousers"
      },
      "pose": "Sitting on the left, relaxed posture, looking towards the screen."
    },
    {
      "id": "daughter",
      "demographics": "Young girl, approximately 6-8 years old",
      "appearance": {
        "hair": "Light brown/blonde, long, straight with bangs",
        "expression": "Smiling, engaged"
      },
      "clothing": {
        "top": "Beige ribbed knit sweater",
        "bottom": "Blue jeans"
      },
      "pose": "Sitting in the center between parents, holding a large glass bowl of popcorn."
    }
  ],
  "environment": {
    "location": "Modern living room interior",
    "furniture": {
      "sofa": "Plush, teal/green velvet sectional or sofa",
      "shelving": "White wooden shelving units with books and decor items in the background"
    },
    "decor": {
      "wall_art": "Abstract framed print in neutral tones",
      "plants": "Green indoor houseplant (Monstera visible) to the left",
      "props": "Woven baskets on top shelves, candles"
    },
    "foreground": "Blurred silhouette of a laptop screen or tablet"
  },
  "lighting_and_color": {
    "lighting_type": "Warm indoor ambient lighting",
    "sources": "Soft lamp light from background, possible glow from the screen in front",
    "color_palette": [
      "Teal",
      "Mustard Yellow",
      "Warm Beige",
      "Denim Blue",
      "Soft White"
    ],
    "tone": "Warm, inviting, 'Hygge' aesthetic"
  },
  "technical_details": {
    "camera_angle": "Eye-level, straight on",
    "composition": "Rule of thirds, centered subjects",
    "depth_of_field": "Shallow (bokeh effect on background shelves and foreground laptop)",
    "focus": "Sharp focus on the family's faces",
    "style": "Lifestyle photography, cinematic realism, stock photography quality"
  }
}

Now lets do nothing to the code but run it through Nano Banana Pro and see what the output is:

Looks different, possibly not as good but it has now reproduced an image we can edit endlessly by altering the code.

Let's ask the scene to change up slightly using the code by altering the camera angle, changing the lady’s hairstyle, putting the guy behind the sofa and altering the little girl’s haircut:

Now you can see the power…