What is a model? Which should I use?

When diving into the world of AI image generation, you're bound to encounter the term 'model'. You might be wondering, what exactly is a model? Which one should I use? We assume you’re reading this because you have those questions, and have broken it down for you in an easy-to-understand way.

What is a Model?

In the context of AI, a 'model' is like a specialized dictionary. But instead of providing definitions to words, it associates words and phrases with the machine’s understanding of what those words look like. When you look up a term such as "a serene beach at sunset", instead of seeing a written explanation you’d see a beautiful, rich image that captures that phrase.

This 'dictionary' is built using a process called machine learning, where a system learns what certain words look like.

When you input a prompt, the AI model refers to its 'dictionary' to try to understand what you want to see, then uses this information to generate a unique image. Each model has a different definition and perspective of what the words you are using should look like, depending on what it has been trained on.

Note that any models labeled as “Inpainting” models are specialist models for use on the Unified Canvas, and will not generate typical results in the Text to Image or Image to Image modes.

Choosing the Right Model

So which model should you use? The answer depends on what you need. There are various models available in Invoke AI, each with its strengths and unique capabilities.

Some models are generalists, like a comprehensive English dictionary. They've been trained on a wide range of data, making them flexible and useful for a broad range of prompts. Models like Deliberate (by Xpuct) and Seek.art's MEGA Model fall into this category.

Other models are specialists, more like a dictionary for a specific field, say, Invoke's 'Fantasy and Art' model. These models have been trained (or 'fine-tuned') on a set of original fantasy art by Zovya, and other community submissions, and they excel at generating images related to those topics. If you're looking to generate images of fantasy RPG artwork, for instance, such a model would be ideal.

When choosing a model, consider your needs. If you're exploring different styles and subjects, a generalist model may be your best bet. But if you're focusing on a specific theme, look for a specialist model in that area.

Test and Learn

Just as you might flip through a few dictionaries to find the one that fits your needs, don't be afraid to try different models to see what works best for you. Test different prompts, observe the results, and adjust accordingly.

In summary, an AI model is like a sophisticated 'dictionary' linking words with images. Choosing the right one involves understanding your specific needs and being open to a bit of exploration and experimentation.