>>106298859
The model shilling is just very inorganic. When Flux came out, everyone acknowledged its issues, from recycled style to plastic skin, and this is an issue that persisted into finetunes and was evident/obvious to anyone using it. But ever since Qwen came out, there has only been praise for it, no criticism whatsoever. Compare that to how much shit Chroma gets in spite of solving the same shit that Qwen solves (better prompt following, etc...), but better. Suddenly everyone is blind to Qwen's own shortcomings not just compared to Chroma, but also as a whole. The model does not generalize styles. It was trained on the same subset of GPT/Seedream slopped styles and that's all it can do out of the box. Qwen LoRAs may fix that, but they can't fix same seeds and lack of creativity. They alone can't teach the model the thousands of artists and mediums that have been lost in training. But just because it's slightly better than Flux dev, the community copes that the model is "good". Now, Wan 2.2 is relatively new tech, and that deserves all the praise it gets, but cmon...