flux.1.dev fp8 version model

V9.0 adds more training data on fashion and product design elements, resulting in more vivid depictions of minimalist or extremely complex designs. As the version name suggests, V9.0 has more vibrant colors and more stable color mixing. V9.0 also stabilizes the output of long sentences in English, just use double quotes and add the desired text inside the quotes.

The model has been updated to 1.1 and has improved its response to different races, with more delicate skin tones and more accurate semantic understanding. The model can generate high-quality realistic images with multiple scenes and stables. The model has also been trained to recognize and respond to elements such as cheongsam and hanfu, which are specific to the East Asian culture.

ComicTrainee-V2.0 large model for comic-style anime illustrations. Improved performance and universality, with better responses to different facial expressions, actions and elements; *Requires ADetailer to be enabled for face and hand repair.

This model combines traditional Chinese elements with modern urban elements, making it suitable for generating novel-style promotional content

We have updated Neta Art XL V2.0, which is based on V1.0 and optimized character poses, further enhanced hand stability, and is very good at telling stories with atmospheric images.

Real Scene Photography XL Elite KV | E-commerce Product Photography Poster Visual Design

A large model with full elements related to cities, trained on a large number of high-resolution effect diagrams and fused by multiple models, with strong generalization ability, suitable for most city-related generation scenarios. For special scenarios, pairing with specialized lora and prompts can achieve better performance.