How to use the new images.ai

Make this, not that

The first box "make this" tells the AI what you want to see, and "not this" excludes the elements that you don't want to see. For example, you might say "make a beautiful painting of the city" but not "factories". The "not this" area is what is known as the negative prompt area. Another example: Make this... a very handsome man, but not this... a muscles, tattoos, beard, mustache, facial hair. You can use commas in both fields to express your ideas, and correct anatomy. Make party balloons, but not pink. Make this... a lovely hand holding an green apple, but not... red, six fingers, four fingers, jewelry, or hairy. Help the AI understand your vision.

Pick the recipe

A recipe is a combination of art styles, prompt templates, strong negative prompts (explained above) and other settings. They work a little like Instagram filters, but are much more powerful. You can choose a recipe and then add other AI concepts to create your own unique look.

What is a Concept, and what is a LoRA?

We are rolling out support for many different kinds of visual AI models, also known as concepts. A concept can be anything from a base art style, to the subject or character in the art, to a pose, or a visual effect that modifies all of the above. You can add multiple concepts to your art, but it works best when there are three or less. There can only be one "base" or "full" model. A LoRA (Low-rank adaptation) is a mini-model that interacts with the base concept. Inpainting, Textual Inversions, and Lycoris are other types of AI models. We are working towards having hundreds of them

What are weights?

Weights tell the AI how much attention to pay to a keyword or concept, like the volume on a DJ turntable when multiple things are in play. For example, a weight of 0.5 dog and 1.0 iguana will make a crazy pet. A normal weight is 1. More than 1 means very much, a decimal greater than zero means a little bit, and zero means off. Less than zero is experimental, and may lead to surreal results -- negatives mean not only off but to reduce the general idea from other concepts... you probably don't want to use negatives in most cases. A safe number for blending weights is between 0.3 - 0.7.

Troubleshooting

When I use concepts, why are my images blue or jaggy?

First, check that your concept has a base model. Using a LoRA type as the base is not recommended, as these are mini models that need the support of a base to produce a clear image. Stacking multiple concepts is a tricky balancing act, where one can overpower the other. Try using the weights system and reducing the number of concepts and the weights to a safer value, and adding others gradually.

Why do I receive the message: “ehh… I don’t think I want to see that”?

Due to ongoing issues with users who are attempting to generate adult images, and in some instances child pornography, we have had to block many words which may by themselves be harmless. Unfortunately, in many instances this results in this message popping up for users who are simply trying to use the platform. We understand this is frustrating and we are working on reducing the number of blocked terms.

Can I use the images for commercial purposes?

You are free to use the images for any purpose including on websites, book covers, PDFs, etc. No attribution is required, but we do appreciate when it is given.

Are there any copyright issues with me using the images for a website?

The images are copyright free.

Do I need to give credit or attribution to your website when using images?

No attribution or credit is required, but we do appreciate when it is given. The more we can grow the community the more we can invest in additional servers and resources.

How many images can I generate?

In order to conserve server resources this is currently set to 25 images per day. Once IMG tokens are launched, IMG token holders will have the capacity for additional renderings depending on the number of tokens that are held (staked). Unlimited renderings will be available for token holders who stake 1000 tokens.

Can you save any rendered image in a bigger format?

This is something that we are reviewing for future versions of Images.ai. At the moment you cannot save the images in a bigger size, one option is to use the HitPaw Image Enhancer service. This is currently the best AI image enhancer available for Windows and Mac. It is simply one-click to automatically enhance image quality and resolution, and it enables you to enlarge and upscale images without quality loss at all.

Are the images / prompts "Unique to me" or freely available for the rest of the community to edit or use?

Currently generated images can be used by anyone in the community regardless of who generates the image. In future versions we may offer the option to buy exclusivity to a generated image in the form of an NFT (Non-fungible Token).

How are images generated?

Our platform currently uses stable diffusion open-source software. This is a type of Generative AI that generates images based on text prompts. The creativity of the prompt results in the image that is rendered.

Do you have an API?

At the moment there is no API. This may change in future versions of the images.ai platform.

How do you monetize?

Version 1.0 of images.ai is focused on growth. Future versions of the website will have an IMG utility token.