Skip to main content
All CollectionsMAGROUND AI
Best practice for trained AI location models
Best practice for trained AI location models
Updated over a week ago

What is location-based image generation?

Our AI-driven platform allows users to create stunning, realistic images by leveraging trained location models or reference images. With location-based image generation, you can either start with a pre-trained location, alter it to suit your needs, or create entirely new scenes based on your own inputs.

How do I get started?

  1. Use a Pre-Trained Location: Our platform comes with models trained on specific locations. These locations serve as a base, allowing you to generate images that capture the environment's key features like weather, time of day, and surface elements (roads, buildings, landscapes). You can start by generating a prompt based on one of these locations.

  2. Customize the Prompt: Once you've selected a base location, you can modify the prompt to better fit your project. Some examples of modifications include:

    • Adjusting the weather conditions (e.g., sunny to overcast).

    • Changing the time of day (e.g., noon to golden hour).

    • Modifying surface elements (e.g., adding or removing trees, altering building structures).

    Tip: Keep the changes subtle if you want the final image to closely resemble the original trained location.

Original location:

Altering the original prompt generated from the original into a night scene:

Result with altered prompt:

Result with original prompt

  1. Upload a Reference Image: If you want more control, you can upload your own reference image. The AI will analyze this image and generate a new prompt based on its features. This allows for greater creative flexibility in your final output.

  2. Start from Scratch: Prefer to start with a blank canvas? Simply input a custom prompt without using a pre-trained location or reference image. This option provides unlimited possibilities for creating unique visuals.

Tips for Best Results:

  • Stay true to the location’s core elements: For more natural-looking images, make only subtle changes to the trained location’s features.

  • Experiment: Play around with different settings like weather or time of day to see how small changes can dramatically affect the final image.

Can I upload my own reference images?

Yes! Uploading a reference image will generate a new prompt based on that image's key features. This method gives you even more flexibility and control over the image generation process.

How can I get the best visual results?

For the most realistic outcomes, avoid making drastic changes to the trained location prompts. Instead, try minimal adjustments to retain the authenticity of the base location while exploring subtle variations.

Can I create images from scratch?

Absolutely! You can generate images from a completely custom prompt, bypassing both the pre-trained locations and reference images. This feature gives you limitless creative options to explore.

Did this answer your question?