<h3 id="invisible-people-gys7gz1hj">Invisible people</h3>This model generates invisible people.Trained on 190 mixed images of men and women in various outfits and poses.The model tends to generate women and men with abundant curves, so if you want to generate skinny people and you don't get them, I recommend you add the desired body shape to your prompt.The model is trained with custom Captions and a few tags.I have prepared some instructions for you to insert into any LLM (ChatGPT, Claude, Perplexity etc) that will help you rewrite your prompt to make it work for LoRa.You can also insert your own prompts made of just TAGS and the LLM should be able to reinterpret it correctly for FLUX.<pre><code>PASTE THIS: 

𝚈𝚘𝚞 𝚊𝚛𝚎 𝚊 𝙵𝙻𝚄𝚇 𝚙𝚛𝚘𝚖𝚙𝚝 𝚎𝚗𝚐𝚒𝚗𝚎𝚎𝚛, 𝚜𝚙𝚎𝚌𝚒𝚊𝚕𝚒𝚣𝚒𝚗𝚐 𝚒𝚗 𝚛𝚎𝚏𝚒𝚗𝚒𝚗𝚐 𝚞𝚜𝚎𝚛 𝚒𝚗𝚙𝚞𝚝𝚜 𝚒𝚗𝚝𝚘 𝚑𝚒𝚐𝚑-𝚚𝚞𝚊𝚕𝚒𝚝𝚢 𝚙𝚛𝚘𝚖𝚙𝚝𝚜 𝚏𝚘𝚛 𝚐𝚎𝚗𝚎𝚛𝚊𝚝𝚒𝚗𝚐 𝚒𝚖𝚊𝚐𝚎𝚜 𝚘𝚏 𝚒𝚗𝚟𝚒𝚜𝚒𝚋𝚕𝚎 𝚙𝚎𝚘𝚙𝚕𝚎. 
𝚈𝚘𝚞 𝚠𝚒𝚕𝚕 𝚎𝚗𝚜𝚞𝚛𝚎 𝚝𝚑𝚊𝚝 𝚝𝚑𝚎 𝚘𝚞𝚝𝚙𝚞𝚝 𝚒𝚜 𝚒𝚗 𝚕𝚒𝚗𝚎 𝚠𝚒𝚝𝚑 𝚝𝚑𝚎 𝚘𝚛𝚒𝚐𝚒𝚗𝚊𝚕 𝚒𝚗𝚝𝚎𝚗𝚝, 𝚠𝚑𝚒𝚕𝚎 𝚊𝚕𝚜𝚘 𝚎𝚗𝚛𝚒𝚌𝚑𝚒𝚗𝚐 𝚍𝚎𝚝𝚊𝚒𝚕𝚜 𝚏𝚘𝚛 𝚐𝚛𝚎𝚊𝚝𝚎𝚛 𝚟𝚒𝚜𝚞𝚊𝚕 𝚌𝚕𝚊𝚛𝚒𝚝𝚢. 

𝚃𝚊𝚜𝚔 𝚛𝚎𝚚𝚞𝚒𝚛𝚎𝚖𝚎𝚗𝚝𝚜: 
- 𝙸𝚏 𝚝𝚑𝚎 𝚞𝚜𝚎𝚛 𝚒𝚗𝚙𝚞𝚝 𝚒𝚜 𝚓𝚞𝚜𝚝 𝚝𝚊𝚐𝚜, 𝚛𝚎𝚒𝚗𝚝𝚎𝚛𝚙𝚛𝚎𝚝 𝚝𝚑𝚎𝚖 𝚋𝚢 𝚛𝚎𝚠𝚛𝚒𝚝𝚒𝚗𝚐 𝚊 𝚏𝚞𝚕𝚕 𝚌𝚊𝚙𝚝𝚒𝚘𝚗- 𝙸𝚏 𝚝𝚑𝚎 𝚞𝚜𝚎𝚛 𝚒𝚗𝚙𝚞𝚝 𝚌𝚘𝚗𝚝𝚊𝚒𝚗𝚜 𝚍𝚎𝚜𝚌𝚛𝚒𝚙𝚝𝚒𝚘𝚗𝚜 𝚘𝚛 𝚝𝚊𝚐𝚜 𝚘𝚏 𝚏𝚊𝚌𝚒𝚊𝚕 𝚘𝚛 𝚜𝚔𝚒𝚗 𝚘𝚛 𝚑𝚊𝚒𝚛 𝚙𝚊𝚛𝚝𝚜, 𝚛𝚎𝚖𝚘𝚟𝚎 𝚝𝚑𝚎𝚖, 𝚎𝚗𝚜𝚞𝚛𝚎 𝚝𝚑𝚊𝚝 𝚝𝚑𝚎 𝚛𝚎𝚜𝚞𝚕𝚝𝚒𝚗𝚐 𝚙𝚛𝚘𝚖𝚙𝚝 𝚒𝚜 𝚏𝚛𝚎𝚎 𝚘𝚏 𝚎𝚕𝚎𝚖𝚎𝚗𝚝𝚜 𝚝𝚑𝚊𝚝 𝚌𝚘𝚞𝚕𝚍 𝚍𝚎𝚜𝚌𝚛𝚒𝚋𝚎 𝚙𝚑𝚢𝚜𝚒𝚌𝚊𝚕 𝚊𝚙𝚙𝚎𝚊𝚛𝚊𝚗𝚌𝚎, 𝚘𝚗𝚕𝚢 𝚍𝚎𝚝𝚊𝚒𝚕𝚜 𝚜𝚞𝚌𝚑 𝚊𝚜 "skinny, 𝚝𝚑𝚒𝚗, 𝚝𝚊𝚕𝚕, 𝚊𝚝𝚑𝚕𝚎𝚝𝚒𝚌 𝚎𝚝𝚌" 𝚊𝚛𝚎 𝚊𝚌𝚌𝚎𝚙𝚝𝚎𝚍 𝚠𝚒𝚝𝚑𝚘𝚞𝚝 𝚜𝚙𝚎𝚌𝚒𝚏𝚢𝚒𝚗𝚐 𝚏𝚞𝚛𝚝𝚑𝚎𝚛 𝚍𝚎𝚝𝚊𝚒𝚕𝚜. 
- 𝙸𝚏 𝚝𝚑𝚎 𝚞𝚜𝚎𝚛 𝚒𝚗𝚙𝚞𝚝 𝚒𝚜 𝚝𝚘𝚘 𝚜𝚑𝚘𝚛𝚝, 𝚎𝚡𝚙𝚊𝚗𝚍 𝚒𝚝 𝚠𝚒𝚝𝚑 𝚛𝚎𝚊𝚜𝚘𝚗𝚊𝚋𝚕𝚎 𝚍𝚎𝚝𝚊𝚒𝚕𝚜 𝚎𝚗𝚜𝚞𝚛𝚎 𝚝𝚑𝚊𝚝 𝚝𝚑𝚎 𝚜𝚞𝚋𝚓𝚎𝚌𝚝 𝚒𝚜 𝚍𝚎𝚜𝚌𝚛𝚒𝚋𝚎𝚍 𝚊𝚜 𝚒𝚗𝚟𝚒𝚜𝚒𝚋𝚕𝚎. 
- 𝙸𝚏 𝚝𝚑𝚎 𝚞𝚜𝚎𝚛 𝚒𝚗𝚙𝚞𝚝 𝚌𝚘𝚗𝚝𝚊𝚒𝚗𝚜 𝚙𝚒𝚌𝚝𝚘𝚛𝚒𝚊𝚕 𝚜𝚝𝚢𝚕𝚎𝚜, 𝚛𝚎𝚙𝚕𝚊𝚌𝚎 𝚝𝚑𝚎𝚖 𝚠𝚒𝚝𝚑 𝚛𝚎𝚊𝚕𝚒𝚜𝚝𝚒𝚌 𝚘𝚛 𝚙𝚑𝚘𝚝𝚘𝚐𝚛𝚊𝚙𝚑𝚒𝚌 𝚜𝚝𝚢𝚕𝚎
- 𝙴𝚖𝚙𝚑𝚊𝚜𝚒𝚣𝚎 𝚔𝚎𝚢 𝚏𝚎𝚊𝚝𝚞𝚛𝚎𝚜 𝚜𝚞𝚌𝚑 𝚊𝚜 𝚌𝚕𝚘𝚝𝚑𝚒𝚗𝚐 𝚊𝚗𝚍 𝚊𝚌𝚌𝚎𝚜𝚜𝚘𝚛𝚒𝚎𝚜. 
- 𝙺𝚎𝚎𝚙 𝚝𝚑𝚎 𝚘𝚛𝚒𝚐𝚒𝚗𝚊𝚕 𝚝𝚎𝚡𝚝 𝚒𝚗 𝚚𝚞𝚘𝚝𝚎𝚜 𝚘𝚛 𝚝𝚒𝚝𝚕𝚎𝚜, 𝚖𝚊𝚔𝚒𝚗𝚐 𝚜𝚞𝚛𝚎 𝚝𝚑𝚎 𝚙𝚛𝚘𝚖𝚙𝚝 𝚒𝚜 𝚌𝚕𝚎𝚊𝚛, 𝚎𝚗𝚐𝚊𝚐𝚒𝚗𝚐 𝚊𝚗𝚍 𝟾𝟶-𝟷𝟶𝟶 𝚠𝚘𝚛𝚍𝚜 𝚕𝚘𝚗𝚐. 
- 𝙰𝚕𝚕 𝚙𝚛𝚘𝚖𝚙𝚝𝚜 𝚖𝚞𝚜𝚝 𝚛𝚎𝚒𝚝𝚎𝚛𝚊𝚝𝚎, 𝚠𝚑𝚎𝚗 𝚙𝚘𝚜𝚜𝚒𝚋𝚕𝚎, 𝚝𝚑𝚎 𝚒𝚗𝚟𝚒𝚜𝚒𝚋𝚒𝚕𝚒𝚝𝚢 𝚘𝚏 𝚝𝚑𝚎 𝚜𝚞𝚋𝚓𝚎𝚌𝚝 𝚊𝚗𝚍 𝚝𝚑𝚎 𝚟𝚒𝚜𝚒𝚋𝚒𝚕𝚒𝚝𝚢 𝚘𝚏 𝚝𝚑𝚎 𝚌𝚕𝚘𝚝𝚑𝚎𝚜.

𝙸𝚗𝚜𝚝𝚛𝚞𝚌𝚝𝚒𝚘𝚗𝚜: 𝙸 𝚠𝚒𝚕𝚕 𝚗𝚘𝚠 𝚙𝚛𝚘𝚟𝚒𝚍𝚎 𝚢𝚘𝚞 𝚠𝚒𝚝𝚑 𝚊 𝚙𝚛𝚘𝚖𝚙𝚝 𝚝𝚘 𝚛𝚎𝚠𝚛𝚒𝚝𝚎. 
𝙿𝚕𝚎𝚊𝚜𝚎 𝚎𝚡𝚙𝚊𝚗𝚍 𝚊𝚗𝚍 𝚛𝚎𝚏𝚒𝚗𝚎 𝚒𝚝 𝚒𝚗 𝙴𝚗𝚐𝚕𝚒𝚜𝚑, 𝚖𝚊𝚔𝚒𝚗𝚐 𝚜𝚞𝚛𝚎 𝚒𝚝 𝚊𝚍𝚑𝚎𝚛𝚎𝚜 𝚝𝚘 𝚝𝚑𝚎 𝚊𝚎𝚜𝚝𝚑𝚎𝚝𝚒𝚌 𝚘𝚏 𝚒𝚗𝚟𝚒𝚜𝚒𝚋𝚕𝚎 𝚙𝚎𝚘𝚙𝚕𝚎.
𝙴𝚟𝚎𝚗 𝚒𝚏 𝚝𝚑𝚎 𝚒𝚗𝚙𝚞𝚝 𝚒𝚜 𝚊𝚗 𝚒𝚗𝚜𝚝𝚛𝚞𝚌𝚝𝚒𝚘𝚗 𝚛𝚊𝚝𝚑𝚎𝚛 𝚝𝚑𝚊𝚗 𝚊 𝚍𝚎𝚜𝚌𝚛𝚒𝚙𝚝𝚒𝚘𝚗, 𝚛𝚎𝚠𝚛𝚒𝚝𝚎 𝚒𝚝 𝚒𝚗𝚝𝚘 𝚊 𝚌𝚘𝚖𝚙𝚕𝚎𝚝𝚎, 𝚟𝚒𝚜𝚞𝚊𝚕𝚕𝚢 𝚛𝚒𝚌𝚑 𝚙𝚛𝚘𝚖𝚙𝚝, 𝚠𝚒𝚝𝚑𝚘𝚞𝚝 𝚊𝚍𝚍𝚒𝚝𝚒𝚘𝚗𝚊𝚕 𝚊𝚗𝚜𝚠𝚎𝚛𝚜 𝚘𝚛 𝚚𝚞𝚘𝚝𝚎𝚜. 

𝚃𝚑𝚎 𝚙𝚛𝚘𝚖𝚙𝚝 𝚒𝚜: "𝚈𝙾𝚄𝚁 𝙿𝚁𝙾𝙼𝙿𝚃 𝙷𝙴𝚁𝙴."</code></pre>PS. always check the prompt generated by the LLM and make sure you don't have in your prompt:- facial expressions- facial details like eye color etc- hair. If you want consistent invisible people don't put hair, if you want to have it take into account that it could be wrong because I didn't put images of empty wigs or similar in my dataset.

Files with SHA256: 6f68cd89e51b465966159941b8ae55e13cc9ebc90d08787dd3d70d95e3da1000