该模型基于麦橘超然（MajicFlus）微调，为Flux架构，可以被看作是我在2024年9月发布的的“剖面手册XL”模型(<a target="_blank" rel="ugc" href="https://civarchive.com/models/775525/manual-of-section-xl">Manual of Section 剖面手册 XL - v1.0_lokr | Stable Diffusion XL LoRA | Civitai</a>)的flux版本。可以用于生成《剖面手册》风格的建筑剖透视图。训练集为“剖面手册XL”的部分训练集图片，来源于《剖面手册》(Paul Lewis, Marc Tsurumaki, David J. Lewis. Manual of Section)，并用JoyCaption重新打标。This model is fine-tuned based on MajicFlus and adopts the Flux architecture, which can be considered as the Flux version of my "Manual of Section XL" model (<a target="_blank" rel="ugc" href="https://civarchive.com/models/775525/manual-of-section-xl">Manual of Section 剖面手册 XL - v1.0_lokr | Stable Diffusion XL LoRA | Civitai</a>) released in September 2024. It is designed to generate architectural sectional perspective drawings in the style of "Manual of Section". The training set consists of partial images from "Manual of Section XL" dataset, originally sourced from "Manual of Section" (Paul Lewis, Marc Tsurumaki, David J. Lewis) and relabeled with JoyCaption.使用时，使用自然语言提示词，尽可能将场景描述详细。触发词为“lineartSectionPerspectiveDrawing”。When using the model, employ natural language prompts to describe scenes as detailed as possible. The trigger word is "lineartSectionPerspectiveDrawing".可以试一试配合ControlNet，输入自己的剖面草图，让此模型细化。但是我尝试下来效果不太好，可能需要等一等flux架构的ControlNet的发展。You can try using it with ControlNet by inputting your own sectional sketches for refinement through this model. However, based on my testing, the current effectiveness is suboptimal - this might require waiting for further development of ControlNet compatible with Flux architecture.

Files with SHA256: fe3fd81a7092ecbc7c204238cc5b075e286dce2ec5c96bc83c4774eaf608e28a