Understanding models on LetzAI
Models on LetzAI let you include a specific person, character, object, or visual style in any generation. Tag one in your prompt, and the AI incorporates it into your image. This guide explains what is happening when you do that.
What is a model?
A model is built from a collection of images, called a dataset. That dataset represents a specific subject: a person's face, a recurring character, a visual style, or a branded object.
Once a model exists, you can tag it in a prompt to include that subject in a generation.
How it works
When you tag a model, LetzAI does the following:
- The platform looks at the images in that model's dataset.
- It automatically selects a few images to use as a reference.
- Those reference images are passed to the AI alongside your prompt.
- The AI uses them to understand what the subject looks like and includes it in your generation.
That is it. No waiting, no extra setup. Under the hood, model tagging is very similar to using reference images — except LetzAI picks the references for you from your model's dataset.
Why this is useful
- Instant results. Models are ready the moment they are created.
- No manual steps. You do not need to upload or select a reference image yourself each time you generate.
- Use community models. You can tag public models made by other LetzAI users without them sending you reference images.
One thing to know
The platform picks the reference images automatically, so you do not choose which ones it uses.
Most of the time this works well. Occasionally, the selected image may not be the best match for your prompt — a different angle, expression, or lighting than you had in mind can affect the result.
If a generation using a model does not look quite right, this is often why. Try generating a few more variations. With different selected reference images, you may get a better outcome.
Creating your own models
You add a model by uploading images that define your subject. That builds the dataset LetzAI draws from when you tag the model. Create or manage models from the Train page, or use Character Studio to define a character without starting from photos.
Tips for better results
When building your own model
- Use clear, well-lit images where the subject is easy to recognize.
- Avoid blurry, cropped, or cluttered images.
- Ask yourself: would I be happy if this image were used as my reference?
When using any model
- Be specific in your prompt about the look you want.
- If the result is not what you expected, generate again before assuming the model is wrong.
- When you want full control, open the Edit tab and manually upload the best possible reference image, then generate from there.