LetzAI V2 Prompt Guide

Our guide for generating high-quality images like a pro.

This guide will teach you everything there is to know to get the best AI Images out of LetzAI. We'll cover the basics, important dos and don’ts, and how to use all our features.

For complete beginners: Image generation works through prompts.
A 'prompt' is a written instruction you give to an artificial intelligence system to generate an image, like asking a friend to draw something based on your description. The quality of your images image depends heavily on your prompt choice.

LetzAI basics and guidelines

Using AI Models in your prompts

Unlike many AI systems, LetzAI focuses on personalized image generation. For example while other AI apps can generate a generic image of a statue, they won't be able to replicate niche objects like Luxembourg's iconic "Gëllefra" statue. Similarly, they can produce images of non-existent humans or famous people, but not of your friends wearing a specific product.

LetzAI does not know the "Gëllefra" statue by default, but users can teach it what a person, object, or style looks like. We call this "Creating an AI Model." Here is the AI Model of said "Gëlle Fra" on LetzAI.

Through AI Models created by the community, LetzAI learns about the unique characteristics of specific objects, people, or art styles.

AI Models can be public or private, which means you get to decide if you want others to be able to use your model or not. If you change your mind, you can publish or unpublish your model at any time.

Tag models with @ to generate images with specific content

By training a new personal AI Model, you’re providing LetzAI with new information that tells it what your object/person/style looks like. Once your model training is complete and you’ve activated your model, you can use that model in your prompt by typing the “@” character in the prompt field, and selecting your AI model.

We call this “tagging” and it works just like on other social media platforms. Whenever you tag and use a model, you instruct our system to apply that model in your prompt.

You can also use models that other users have created and set to “public”. When you type "@" in the prompt box, a list will pop up, showing you all the AI models you can use. This list contains available AI models, either because they are your privately trained models, or because other users publicly provided them.



Important: You cannot tag a model that doesn’t exist in this list. For example, if you are trying to use the tag @YourName but this model does not exist, LetzAI won’t know what you want to generate, hence your prompt will have no effect.

In other words, if a model doesn’t exist in the public database you’ll need to train that model yourself.

Use simple scenarios to get more coherent results

You usually get more coherent results with simple situations.

For example, the prompt 'Labradoodle sitting in front of a tree in a green park during summer' tends to work better than 'Labradoodle jumping and barking around a tree, chasing a squirrel sitting on a branch and looking down'.

Add details, but don’t overdo it

The more detailed and complex your instructions are, the more you steer the model's creative process. However, this also increases the possibility that the outcome may deviate from your specifications.

If you have a specific idea, be careful not to add too many details to your prompt, as this will increase the chance of your results being more chaotic.

Illustrations vs. Photorealism

You should always indicate style-related keywords to specify the style you want your image to have.

If you want a photo, make sure to add keywords like “photography” to your prompt. If you want an illustration, add keywords describing the illustration style you’re looking for.

To increase your success rate for a style, you should add these words in the beginning of your prompt.
Take this example of our CEO Misch Strotz:

Adding your own Art Style

If you're an artist, and want to add your own style to LetzAI, you can do by training your own Style model. Keep in mind, that adding keywords to your prompts will also help in getting your style right! So if your art style is illustrative, consider adding words like "illustration" as a type to your model, or inside of your prompts.

You can find a complete list of available Style models on the model page here on the Explore page.

To get even more inspiration, you can also open community images on the feed to see which models others have used for their images and use them with your own prompt.

If you’re interested in training your own models, you can find a complete guide here.

Limit Model Combinations

Currently, combining multiple models in one prompt can pose certain challenges.

For example, using two different person models in one image will likely merge their faces and can result in bad-quality images.

If you want to mix models, it can help to reduce their intensity to make them blend and harmonize better. In the second part of this article, under 'Model intensity adjustments', you’ll find tips on how to customize the strength of each model.

Avoid negative phrasing

Using negative phrasing can lead to misunderstandings, resulting in images with the very elements you wished to exclude. So don’t include things in your prompt that you don’t want to see.

For example, don’t prompt “A 30-year-old man on the beach without a surfboard”, or it might do this:

Prompt features and settings

LetzAI also offers a range of settings and features to make your life easier. Let’s briefly go over the most important ones:



LetzAI also offers a range of settings and features to make your life easier. Let’s briefly go over the most important ones:

Generation settings

Clicking the gear icon opens the generation settings with various adjustment options.



1. Dimension settings



Clicking the gear icon opens the generation settings with various adjustment options.

2. Aesthetics options: Quality & Creativity



Clicking the gear icon opens the generation settings with various adjustment options.

The Quality and Creativity sliders range from 1 to 5, with 2 being the default setting.

Quality makes the image crisper and sharper, improving overall quality. However, it takes more time and uses more credits due to a more intensive creation process.

Creativity tells the system to “deviate” from the original prompt to make images more creative, chaotic, colorful, or exaggerated, depending on your initial input.

Setting Model intensities

The intensity defines how strongly a model behaves inside prompt.

If your intensity is too low, your subject won’t look like the subject you trained. If your intensity is too high, LetzAI will “force” your model on the generation, which can result in bad quality.
By default, the intensity is pre-set for you but sometimes, you may want to adjust a model's intensity manually, for example when using multiple models at once.
Using multiple models usually works best when you use 1 person/object model and apply a second style model.

For example, let’s say we want to generate an image of @mischstrotz in the style of @juliewdesign_retrofilter.

Have a look at what happens when we use two models without changing the intensity:




The style model is conflicting with the @mischstrotz model, causing distortions and little artifacts in the image.

To fix this, we can reduce the intensity of the style model.

Whenever you add a model to your prompt a new slider will pop up and allows you to adjust the model intensity. If you add multiple models, you will also see multiple sliders.



The default intensity is 0.8, which can either be lowered all the way to 0 or increased to a 1.2 value. Below 0.5 the influence of a model will be very small, and anything above 1.0 is only recommended if you really want the content of that model to be very dominant.

In this case, we want to manually define the style model intensity of 0.5, so we add “:0.5” at the end of the model name in our prompt:




As you can see in the image above, turning down the intensity of the style model fixed the face and the artifacts.

The same logic can be applied the other way around.

To put the focus on the style model, we would decrease the intensity of the person model.

Let’s combine the “person” model of @mischstrotz with the classic @mickeymouse1928 digital art-style model for testing purposes.




As you can see, using the default intensities gives us an image that blends cartoon and realistic elements since both models are applied with equal “force”. To achieve a more stylized and cartoonish appearance, we can lower the intensity of Misch's model by adding a lower weight to the model in the prompt:




Decreasing the intensity of Misch’s model while keeping the default intensity for the style model, balanced both models and results in a more stylized output.

Find inspiration on the feed

LetzAI's community feed on the home page is full of great prompts! The combinations of possible prompts are endless, and the feed is a great place to find inspiration for your own prompts. You can also see which models other users have used for their images. If you find a prompt that you like, you can always mix and match it with your own words and see what you get.

Experiment

Remember, these are just guidelines. They should not keep you from experimenting and trying new things. LetzAI is designed to be used creatively. Embrace the unexpected, and don’t worry if things turn out differently than planned. Tweak your prompt and try again.

Keep in mind that results for the same prompt can vary drastically. So don't give up after the first try. This is also true for creating your own models. If something doesn’t come out the way you want, try to optimize your model in your model settings or create a new one.

To learn what to consider when creating your own models, check out this guide.

Congratulations, you’re now all set to get prompting on LetzAI!

And don’t forget to publish your results for the community to enjoy :)

Follow us on Twitter/X

The best way to keep up with everything LetzAI is on Twitter/X.