What is style transfer?
Style transfer uses LoRA-based training on a curated dataset (10+ consistent images) to encode your brand’s visual system. The adapter learns anatomy and proportions, materials, textures and fine detail, color palette and lighting, composition, camera cues, and illustration traits like line weight, brushwork, shading, shape language, and rendering style.
Applied to new prompts, it reproduces those cues consistently across scenes – unlike generic filters. The trained style is yours, reusable with any prompt.
What can you do with it?
Photography – consistent lighting, colour, and treatment across product, lifestyle, and portrait shots. Output: .jpg images.
Illustration – preserve line weight, brushwork, shading, and shape language across scenes. Output: .jpg (vectorization available in the full version).
3D assets – lock a render style (lighting, tone-mapping, post-fx) for turntables and stills. Output: .jpg.
How style training works
A quick guide to train your style, test the 7-day demo, and buy a style slot for full access.
Prepare your dataset
Upload 10+ consistent images that reflect your visual identity. Include variety (people, objects, environments) while keeping the same look & treatment.
Train your style
Your first training is free; each additional training costs 16 credits. Training typically completes in 1–3 hours for all plans
Get a 7-day demo
After training, your style is available as a free 7-day demo. Demo outputs are watermarked and downloads are disabled – perfect for testing prompts and checking fidelity.
Buy the full version (style slot)
Purchase a style slot to convert the demo into the full version. Full version benefits: no watermarks, downloads up to 8K, and commercial rights. You keep ownership of the style. You can share access with your team without exposing the dataset.
Privacy & security
Your training images are stored securely and are not used to train global ai. Sharing a style does not share your dataset.
Updated
30 July 2025
