Detail Master Large 1.0
During the training of the model, we focused on highlighting the finest details. It was created using a special training technology, resulting in a compressed source file of over 5GB, with the uncompressed version reaching 45GB. We utilized and manually tagged over 4 million images for training, a process that began with version b2.0 almost a year ago. Skin tones, lighting, styles, camera types, and much more were added to make this the most detail-enhancing model ever created. The base Lora version reached a VRAM consumption of at least 24GB, but the optimized version now performs excellently with a minimum of 8GB VRAM. For the training process, we used an AMD Instinct™ MI300X graphics module with 192GB of VRAM, utilizing 80% of its capacity. Thank you for the extensive testing and positive feedback on previous versions. We hope you’ll enjoy using this version just as much as the others. Additionally, public testing of the Flux alpha version has also begun, and we warmly welcome everyone to join!
