Custom AI models are trained on your images to learn specific styles, people, objects, or scenes. Once trained, you can use these models to generate new images that incorporate what the AI has learned.
Samsa offers four types of custom models:
Person Models
Train on photos of a specific person to generate new images featuring them in any scenario, style, or setting.
Style Models
Capture a unique visual style—artistic techniques, color palettes, or aesthetic approaches—to apply to any subject.
Object Models
Teach the AI to recognize and recreate specific products, items, or objects with accurate details.
Scene Models
Train on environments, locations, or backgrounds to generate consistent settings for your images.
Upload training images — you only need a minimum of 1 image, but we recommend uploading several for more variety in your results
Choose model type based on what you're training
Start training — it takes around 20 seconds
Use your model — open the dropdown below the prompt field, select your model, and it's added to your prompt like a building block
You'll get a notification when training is complete. Once trained, you can reuse your model unlimited times.
Model training is free and unlimited on all plans. Using your trained models in generation costs the same as regular generation.
Image order matters: If you're not satisfied with results, try reordering your training images
Variety matters: Include different angles, lighting, and contexts
Consistent subject: Ensure all images feature the same subject clearly
High resolution: Better training images lead to better results
You can add custom instructions to your models — rules that the AI follows in all generations using that model. This is great for corporate guidelines or brand rules. Learn more in the Managing Your Models article.
Your trained models are stored in the Models section where you can:
View all your custom models
See training status
Add custom instructions
Share models with your team
Delete models you no longer need
