Train consistent Characters with AI

Our consistent AI Character tool, the Model Trainer, uses a process called DreamBooth. To put it simply, we add a special “instance prompt” token, linked to your unique digital avatar, to the FLUX.1 [dev] image generation model.

Once the Character training is done, you can enable it in the Image Generator, and the system will automatically invoke that token and apply your Character’s traits to every image—no need to add anything extra to the prompt.

You must own the rights to all the pictures you use for training. If the images feature a real person, you’ll need their explicit permission. We have a strict policy against deepfakes and unauthorized usage. Violating these terms may result in a ban or legal action.

Not at all! The process is simple and user-friendly. All you need is a set of diverse photos showing the person from different angles, in various settings and expressions. Each image should have a minimum resolution of 1 Mpx (roughly 1024 x 1024 px). If needed, use our built-in AI Upscaler to quickly increase resolution without losing quality.

The possibilities are endless! Just to name a few, your AI Character can be an invaluable tool for storyboarding and concept art, helping you visualize your ideas with consistent visual representation across different scenes and settings. Thanks to the versatility of the FLUX.1 model, your Character can adapt to any artistic style—whether it's realistic, fantasy, or something entirely unique.

And if you're preparing online courses or tutorials, a digital avatar could appear in all your learning materials, making your content more engaging and professional. Not to mention, the Model Trainer lets you create a virtual you, whether for work purposes or just for fun, or a brand ambassador for your company.