Generate Images with AI

Welcome back. On my screen, you can already see a 3D model I created from an image, and that's exactly what we'll do in this lesson. I'll show you how I generate models for games. Let's get started. Our 3D models will be generated directly from images, so first, we'll need a source image. You can create one yourself or find inspiration online, using tools like Perplexity. Just be careful—much of the concept art online is copyrighted or licensed, so don't directly copy it. While most AI-generated images aren't truly original, they can help you find ideas.

Good sites for inspiration include ArtStation or Pinterest. Let's try Pinterest and search for something like "sci-fi fighter ship." For example, we might use an X-Wing from Star Wars. Copy this image as our starting point, then head over to ChatGPT. Insert your image there, and ask for a clear description, especially focusing on the spaceship. I'm using GPT-4 for this step. Initially, it will likely describe the image specifically (like an X-Wing), but you can ask ChatGPT for a more general description. Then, customize it—change the colors from red to yellow or modify the wings from dual to single-wing. You can either adapt a source image or completely describe a new spaceship yourself.

Now, copy this final description. For generating the actual image, I prefer Midjourney, but you can use any image generator, including fal.ai. Paste the description into Midjourney. For the settings, I choose square mode, raw format, and transparent background. Make sure your description only includes the spaceship itself, removing unnecessary background details.

Generate the image a few times to explore different results. Some might not look right, but others will be great. Pick your favorite design and use Midjourney’s variation tools if you want to adjust it further. Once satisfied, download your chosen image. You can upscale it for better quality if necessary.

Finally, let's take our generated image to fal.ai and create a 3D model from it.