Looking to understand how to generate images? We'll provide some tips and tricks below.
Overview of how it works
Our AI has learnt from millions of images and text all over the web and gained an understanding on how to read and create relevant pictures. Think of it like a 5 year old child's brain. Each time you hit that Generate button, it doesn't copy or pull images from anywhere, but considers the concepts that it knows, and creates new images from scratch each time.
When you start off in the image generator, 'Guide me' is turned on by default to help you build an image description that will give you great results. Here's how to make the most out
- In the 'Describe your desired image' field, specify what subject (i.e. what you want appearing in your image)
- You can then choose from the different fields to define exactly what visual styleyou want from your image.
- Although each field is optional, more details work best if you have something in mind on how you want your image to turn out—as the AI has a huge knowledge of images across the internet, if you don't specify what you want, it'll try to guess, sometimes with variable results.
Here's what each field means:
The general 'medium' of your image. If you want a photorealistic image, choose 'Photograph'. If you want an illustration, choose 'digital art illustration' and for a 3D rendering, chose '3D render'.
The genre of your image, which incorporates values and motifs of a certain aesthetic.
For example, 'cyberpunk' images often make use neon lighting, grimy streets contrasted with high-rise, sophisticated buildings.
Mood and lighting
This contributes the most to how your piece 'feels'. Select as many as you want here. If you want something dark, choose 'stormy' and 'dramatic backlighting'.
Artist or style
You can also just produce images similar to a style of art in a period, or a particular artist.
These tell the AI to produce higher quality output, as the AI has seen a range of work from good to bad.
How to get better results
- Be descriptive: The more detail our AI has to work with, the more specific your results will be. For best results, we suggest having at least a 100 character description in total.
- Avoid conflicting styles: For example, it's hard to convey a 'haunting' mood when there's 'colorful lighting'. Review your inputs to see if there's anything that could be confusing the AI.
- Use multiple modifiers: They might be a 'fingers-crossed' type of input, but it typically helps to tell the AI to produce high quality output by using quality modifiers, or where the art might be featured on (eg. 'on Artstation', a popular portfolio site with beautiful work from around the world).
- Pro tip: Use 'realistic proportions' when generating human-related images, as it helps the AI to proportion humans correctly.
Guidelines for use
We do not accept any use of our AI for irresponsible means, to avoid harming individuals. This includes, but is not limited to:
- Creating hateful, degrading, or otherwise harmful representations of people or their environments, cultures, religions, etc
- Encouraging discriminatory content or harmful stereotypes. Pretending to be someone without their consent.
- Spreading misinformation or fake news.
- Sharing representations of excessive violence and gore.
- Sharing content that is an unauthorized copy of copyrighted or licensed material.
- Sharing content that is an unauthorized alteration of copyrighted or licensed material.
2. Why do faces look so weird?
Right now, our image generator isn't great at generating faces. We're working on this. It turns out that humans are a lot more sensitive to discrepancies in faces, and it's a lot more challenging for AI to understand this.
3. Is the copyright to the images mine?
The images you create are public domain, though you do not own the copyright. They can be used by anyone for any purpose, including commercial purposes. Feel free to use them for anything you like!
4. Why are images blank or black?