AndroFlux is a LoRA model made for FLUX.1 dev by Black Forest Labs, because there’s a glaring lack of accurate male anatomy in AI models. I’m trying to fix that by making sure AI can depict male features more accurately, balancing out the overwhelming female-centric content you see everywhere.Community and SupportJoin The Bulge Discord at <a target="_blank" rel="ugc" href="https://thebulge.xyz">https://thebulge.xyz</a> if you’re into gay AI art—it’s a great place to share and learn. If you want to support the project, whether with feedback or donations, it’s much appreciated. Current StatusThe model isn’t perfect, so don’t expect flawless results.PurposeAndroFlux is all about making sure AI models can represent male anatomy—especially penises—accurately. With that said, it also has a general photographic aesthetic and focuses on twinks so it is still useful for other SFW purposes. It’s a step towards more balanced AI-generated art.Technical DetailsAndroFlux was built and tested primarily on FLUX.1 dev, but it should also be compatible with the Schnell base models, including the quantized fp8 version. These are distilled models, making training tricky and affecting the LoRA’s performance, requiring a lot of optimization.Training Process<ul><li>Resolution Journey: I started training at 1024px, but the model wasn’t converging, so I dropped to 512px. The quality wasn’t great, so I ended up at 768px. This resolution offers a good balance between detail and training time.</li><li>Captioning: Captioning was crucial in training. Contributions came from the AndroAtlas dataset, part of the AndroFlow project. <a target="_new" rel="ugc" href="https://civarchive.com/user/ephermeron01">Ephemeron01</a> from The Bulge Discord assisted with captions using models like llava-hf/llava-v1.6-mistral-7b-hf and a custom-trained PaliGemma VLM. Specific prompts like "man, penis" and the activator "AndroFlux" were also used in later versions.</li><li>Acknowledgments: Huge thanks to bghira and Terminus Research Group for their support, including providing an A100 for some of the training runs.</li></ul>Usage Tips<ul><li>For best results, use prompts that specifically mention male frontal nudity, such as "penis."</li><li>Recommended settings:<ul><li>Resolution: 1024x1024 or 832x1216</li><li>Flux Guidance: 3.5</li><li>LoRA Weight: Start at 1.0, reduce to 0.5 if artifacts appear</li><li>CFG: 2.0 (1.0 will be faster though)</li></ul></li></ul>Suggested Prompts<ul><li>A nude man in his {late teens | early twenties | early thirties}</li><li>His penis is {erect | flaccid}</li><li>His penis is {large | small | average}-sized</li></ul>Version HistorySee "About this version" on the right for version-specific updates.

AndroFlux

Changelog from v8 Configuration<ul><li>MAX_NUM_STEPS: <s>10000</s> &gt;&gt; 30000</li><li>CHECKPOINTING_STEPS: <s>100</s> &gt;&gt; 200</li><li>TRAIN_BATCH_SIZE: 2 (unchanged)</li><li>GRADIENT_ACCUMULATION_STEPS: <s>1</s> &gt;&gt; 2</li><li>RESOLUTION: <s>1024</s> &gt;&gt; 512</li><li>LEARNING_RATE: 1.0 (unchanged)</li><li>OPTIMIZER: Prodigy (unchanged)</li><li>VALIDATION_GUIDANCE: 3.5 (unchanged)</li><li>VALIDATION_GUIDANCE_REAL: <s>1.0</s> &gt;&gt; 2.0</li><li>INSTANCE_PROMPT: Not Set &gt;&gt; "AndroFlux, "</li><li>TRAINER_EXTRA_ARGS:<ul><li>base_model_precision: fp8-quanto (unchanged)</li><li>lora_rank: <s>8</s> &gt;&gt; 16</li><li>max_grad_norm: <s>0.01</s> &gt;&gt; 1.0</li><li>gradient_precision: Not Set &gt;&gt; unmodified</li><li>base_model_default_dtype: <s>bf16</s> &gt;&gt; fp32</li></ul></li></ul>

Files with SHA256: 389455d77dea6b86d1c08e5a36a53b0a603176446e31909bce008889cc4bb2bf