I wanted to get some shoes in for more consistent looks. Got to say, never realise how difficult it is to get consistent shoes. The Lora was trained using 51 images of a pair of Vans look alike shoes, however, depending on which Illustrious checkpoint you use, the results vary, but you still get similar shoes per output. The one checkpoint that gave me the best result so far is <a rel="ugc" href="https://civarchive.com/models/1203050/fabricated-xl">Fabricated XL</a>.<h1 id="currently-the-lora-works-best-using-the-following-settings:">Currently, the Lora works best using the following settings:</h1>Trigger word: CHRNVWhitelowtopAdditional / useful prompts (positive): white low-top sneaker, eyelets, laces(*if you want to add more details, try eyelets, laces)Additional / useful prompts (negative): So far, I didn't use any specific ones for the shoes.Weightage: 0.5 ~ 1(This is the hardest part, I trained only shoes without any other subjects, humans, hands, feet etc., to ensure that the trained model would focus on the shoes, nevertheless, some checkpoints give varied output from weightage 0.5 (sometimes 0.4) ~ 1)Checkpoint Tested: <a target="_blank" rel="ugc" href="https://civarchive.com/models/1203050/fabricated-xl">Fabricated XL</a>, <a rel="ugc" href="https://civarchive.com/models/827184?modelVersionId=2167369">WAI-NSFW-illustrious-SDXL</a>, <a rel="ugc" href="https://civarchive.com/models/1277670?modelVersionId=1896532">JANKU v5 NSFW Trained + NoobAI + RouWei Illustrious XL</a>, <a rel="ugc" href="https://civarchive.com/models/967405/nova-orange-xl">Nova Orange XL</a>(it's the only one I've used so far to test the outputs)<h1 id="weaknesses">Weaknesses</h1>Overall, I've been able to get similar results of the white shoes I want across the images I've been able to generate, though I can't seem to get an exact look alike. Do tailor or adjust the prompts to see if you can get the kind of look that you want, especially when trying to generate subjects with a white pair of shoes. I hit a wall as to trying to create a better Lora since my previous shoes Lora have also not produced decent results.If you have any feedback or suggestions, please let me know. You can download the training model and the pictures to see my workflow.<h1 id="conclusion">Conclusion</h1>I'm quite happy with the decent outputs I get, but more improvement is certainly needed. I'll continue to find ways hopefully, and... try to get better. In the meantime, I'll try to polish my skills before the next set of Lora.I hope you guys will have a good time generating!

Files with SHA256: 4025f4193c24bd2c582e58c5f2f71c800761bf5594728bdf86a0abd93f46e245