Maria Arreghini

LORA
Original


Updated:

This LoRA model is inspired by Maria Arreghini, one of the most dynamic and captivating figures in the Italian sports media scene. Born on April 12, 1998, in Lecco, Maria began her career as a model, winning the title of Miss San Colombano in 2017. Her passion for sports and communication led her to become a prominent sports journalist, hosting shows like Mary’s Place and Rally Dreamer on Sportitalia.

With over 700,000 followers on Instagram, Maria is also a successful influencer, known for her charismatic presence and distinctive style. This LoRA faithfully captures her physical features and graceful demeanor, making it a unique tribute to her multifaceted career.

If you like my models, support my work by giving the model a Star⭐, sharing the images you’ve created while linking the model 📸, and following me if you’d like ❤

For each LoRA, I collect a large set of images to accurately represent the character, ensuring that the model works in all aspects. I work on each image to remove any watermarks, logos, imperfections, and other disturbances, resizing them correctly and making all necessary adjustments to ensure they are ready for training. This process requires great care and attention to detail. Additionally, the model training, whether done locally or online, incurs costs. The entire process of data preparation and model training typically takes around 10 hours. Considering the data preparation, training, and any necessary adjustments, if you decide to support me, know that your contribution will motivate me to continue creating high-quality LoRA, optimized and increasingly refined, to provide the best for those who appreciate my work. If you haven’t found the LoRA you're looking for so far, let me know. If I can, I’ll create it for you.

Version Detail

FLUX.1
9850
10
This model was trained using 985 total images, derived from an original dataset of 197 images, each repeated 5 times. The training ran for 10 epochs, totaling 9850 steps, using bf16 precision and the AdamW8bit optimizer. With a resolution of 1024x1024, support for gradient checkpointing, latent caching, and noise offset, the model ensures a solid balance between visual quality and subject accuracy. It is built on the Flux 1.0 architecture, with dim=16 and alpha=1, optimized for sharp and coherent generations.

Project Permissions

Reprinting is strictly prohibited

    Use Permissions

  • Use in TENSOR Online

  • As a online training base model on TENSOR

  • Use without crediting me

  • Share merges of this model

  • Use different permissions on merges

    Commercial Use

  • Sell generated contents

  • Use on generation services

  • Sell this model or merges

Related Posts