Content
summary Summary

German AI startup Black Forest Labs has unveiled a new API that lets users customize their FLUX Pro AI image model to match specific brand styles using as few as five sample images.

Ad

The company's Finetuning API is meant to address a key limitation of current AI image generators: creating images that accurately match specific visual styles and brand identities. According to Black Forest Labs, the model maintains its flexibility after finetuning while incorporating what it learns from user-provided content into new image creations. The system can generate images at resolutions up to four megapixels.

Black Forest Labs designed the Finetuning API to work across its entire product range. The customized models integrate with both main offerings—FLUX.1 Pro and FLUX1.1 Pro Ultra—as well as specialized tools like FLUX.1 Fill for inpainting and FLUX.1 Depth for structural control.

German media company Burda Verlag uses the Flux Finetuning API to create customized versions of the image model for its various brands. For example, the company's creative teams can now generate images for the Lissy PONY children's brand in minutes, while preserving the brand's distinctive visual elements.

Ad
Ad
Four AI-generated illustrations of a unicorn in the Lissy PONY brand style: as street art, in a fantasy world, with robots and on a sea wave.
Burda Verlag uses a finetuned Flux model to create different image variations for its children's brand Lissy PONY in a matter of minutes. | Image: via Black Forest Labs

Technical details and availability

The API accepts common image formats including JPG, PNG, and WebP files. Users can upload between 1 and 20 training images, each with a maximum resolution of one megapixel. Black Forest Labs says optimal results require at least five high-quality images with clearly defined subjects.

Developers can adjust several training parameters, including the training mode (character, product, style, or general), number of training iterations (minimum 100, default 300), and learning rate. For output, developers can choose between quick training with faster inference or longer processing times that produce higher quality results.

Black Forest Labs has published specific recommendations for different use cases. For example, when training character models, the company suggests using images that show only one figure at a time. For style transfer work, testing higher finetuning strengths typically yields better results.

The FLUX Pro Finetuning API remains in beta testing and currently requires an API key for access. Black Forest Labs hasn't revealed pricing details or announced when the service will become widely available.

Ad
Ad
Join our community
Join the DECODER community on Discord, Reddit or Twitter - we can't wait to meet you.
Support our independent, free-access reporting. Any contribution helps and secures our future. Support now:
Bank transfer
Summary
  • Black Forest Labs has introduced an API that allows the FLUX Pro AI image model to be customized to specific brands and styles using as few as five sample images, addressing a major limitation of current image models in generating precise visual styles and brand identities.
  • According to Black Forest Labs, the model retains its versatility during fine-tuning, but can incorporate learned knowledge about custom content into new image compositions.
  • Burda Verlag uses the API to develop customized versions of the image model for its various brands. For the Lissy PONY children's brand, creative teams can create brand-specific images in minutes while maintaining the brand's distinctive visual elements.
Online journalist Matthias is the co-founder and publisher of THE DECODER. He believes that artificial intelligence will fundamentally change the relationship between humans and computers.
Join our community
Join the DECODER community on Discord, Reddit or Twitter - we can't wait to meet you.