Are you an indie game developer struggling with creating the assets for your game? From characters and environments, to sound effects and music, the process of asset creation can be overwhelming. In this article, we are going to show you how an indie game developer, Alejandro, leverages AI to supercharge his visual asset creation for his games.

As you can see in the image above, Alejandro generated all of these game assets using AI technology, specifically OpenArt’s Photo Booth. This tool allows you to train your own personalized AI models to create visual assets in your desired style. Alejandro, a student, indie game developer, and technologist based in Peru, has participated in the design and development of over 10 games and is recognized as Artificial Artist on OpenArt.

AI as a game asset engine

“Before discovering OpenArt and Photo Booth, I used to get game assets from online markets like Unity Asset Store and Humble Bundle. While these markets are great sources of assets, it’s sometimes hard to find the exact asset I want, because I cannot customize them. I often ended up accumulating a lot of assets that I didn’t use and that took up space in my digital account…“

Upon discovering OpenArt Photo Booth, Alejandro found a good use for all those assets he bought in the past – he has been training many models of icons, characters, potions, weapons, armor, crystals, and more. For example, by using only 10 images of potion images, he’s able to train a model that can consistently generate high-quality images of potion bottles.

Customizable

One of the benefits of using AI to generate game assets is the ability to customize the assets to your exact specifications. With OpenArt’s Photo Booth, he can easily change the colors of the potion, the shape of the bottle, and any other visual elements as desired. The real bonus part is that sometimes the AI can produce unexpected and pleasantly surprising results, adding an element of surprise and creativity to the process.

Time & Cost Efficient

Alejandro has found the AI-generated image method to be faster, more efficient, and less costly compared to other methods, and is excited to be able to do it all himself. He feels as though he has a powerful game asset engine at his disposal that helps to supercharge his game design and development. Overall, the use of AI technology has greatly streamlined and improved his asset creation process.

Quality

When asked about the quality of the AI images, he said that overall he’s really impressed by the quality of AI-generated images and has seen a lot of improvement in recent years. The item icon models are particularly useful to him right now. He often uses the generated images as a starting point and edits them in Adobe Photoshop to adjust any details that don’t quite meet his standards. Then, he will use tools like img2img to further enhance the quality of the icons. The faces generated by AI still need some extra work.

Alejandro has shared many assets he generated on his OpenArt profile so that other developers and designers can use them for free.

Model training tips

Alejandro shares that his process for training models is quite simple and only involves four steps:

1. I define the model I want to train.

2. I then prepare and select the images I’ll use for training.

3. I upload those images to Photo Booth for training and wait about 45 minutes.

4. Finally, I evaluate the model. To give more details, to train a face I need at least 30 high-quality images in 512×512 pixels and apply 6000 steps in the training. Following a similar system, if I don’t exceed 50 images, I keep 200 steps per image. However, if I have more images, I try to reduce the number of steps to 50 or keep the number of steps as a multiple of the number of images used to train the model. For example, with 60 images, I would use 9000 steps.

Another key tip to keep in mind when using AI to generate game assets is the importance of consistency. The more consistent your training images are, the better quality images you will be able to generate with the AI. It’s best to use images that have a similar style, structure, and background in order to achieve the most cohesive and high-quality results. Alejandro found this to be true when he conducted an experiment by training two potion models: one model using images from various sources and another model using images from a single source. He found that the latter model generated higher-quality results more consistently.

The optimistic future

“I would definitely recommend a similar process to non-artist game developers like myself, taking advantage of the opportunities that this new technology brings.”

Alejandro believes that AI technologies will become increasingly sophisticated and able to generate a wider range of game assets, such as character models, icons, and environments. He thinks that this will be particularly useful for indie developers who may not have the resources or time to create these assets manually.

Alejandro is one of the style package creators on OpenArt. A style package (or a preset) is a set of highly-curated prompts that can be applied to a personalized AI model. This saves prompting work to get high-quality images. His Anime & Art style packages are very well-received in the OpenArt community. Creators get revenue shares from the style package sales. If you’re interested in this monetization opportunity as a prompt creator, you can apply for our Preset Creator Program here.

Don’t miss out on the opportunity to supercharge your workflow by training your first model on OpenArt Photo Booth. Start creating high-quality, customizable game assets with ease and efficiency today!