⚽️ Multi-Stage Generative Upscaler (LoRA weights)

Author(s) Luca Martini et al. University of Genova

These weights are a LoRA fine-tune for Black-Forest-Labs/FLUX.1-dev.
They specialise the base model for extreme 64 × 64 ➜ 1024 × 1024 up-scaling (together with JasperAI Control-NET) of football broadcast cut-outs while minimising hallucinations.
The LoRA was trained on 460 high-resolution handcrafted dataset and is meant to be plugged into any Diffusers pipeline that already uses FLUX.1-dev.

WIP

Downloads last month
23
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support