Here's a LoRA I made for The Betrayer (aka. Valen or Commander Valen). Valen is a character that appears in Doom Eternal and Doom Eternal DLC. He's once renowed and respected Sentinel. But after he has dark trade with the Maykrs with exchange to resurrect his own son, other Night Sentinel marked him as betrayer. Hence of his name, The Betrayer. Alas, instead of getting resurrect his son, Maykrs change the son into Icon of Sin form.In Doom Eternal, you can find Betrayer in depth of hell and saying he's worthy in hell. But in Doom Eternal: Ancient of God, he re-appears during Demon and Human war as act of his redemption.Enjoy making artwork of him. Feel free to reviewe the model for my next improvement.Trigger Words:<pre><code>the_betrayer</code></pre>Settings I Used:<ul><li>Resolution: 768x768 and 900x900</li><li>CFG Scale: 7</li><li>Sampling Steps: 20 - 27</li><li>Weight Model: 0.75 - 0.9</li><li>DPM++ 2M SDE Karras</li></ul>Notes:<ul><li><s>Currently, this model only support 2 outfits, Full Armor and Partial Armor that his arms exposed. NSFW and other clothing isn't supported yet. I'm still working for this update since generating synthetic data of Valen is extremely difficult and lack of variety images circulated online.</s> This added support in V1.2</li><li>If you found out that Valen's eye is kinda weird in color---colored white and brown other side---well it is. Apparently, Valen is partially blind canon. Idk if it catharact or what, but he has different eye.</li></ul>If you like my LoRA, you can buy me some coffee:<a target="_blank" rel="ugc" href="https://www.buymeacoffee.com/doomguy1111">https://www.buymeacoffee.com/doomguy1111</a>

Commander Valen (The Betrayer) | DOOM

I updated this model in order to achieve flexibility of this model. As I previously state before, this model is almost impossible to make flexible due to lack of dataset (heck even R34 isn't well rendered because The Betrayer R34 didn't existed, yet). So, I experimented with incorporating data science and data mining approach in order to reach some of support of clothing and R34. Well... this model supported for this. Here's some update changes:<ul><li> Decreased image usage for dataset from 60+ to 50+<ul><li>Originally, I bumped up to 70+ images for this update. Upon inspection, this create dataset imbalance that could result undoing from what I want</li><li>Many of images on this model using kinda same-y style and blurred as hell. I use Valen/The Betrayer in-game screenshot too much that I need to remove it</li><li>Added some high quality and much better clarity for better char accuracy</li></ul></li><li>Fixing model mislabel during labeling<ul><li>I found out that his scarf didn't labelled at all along with his armor</li></ul></li><li>Decreased resolution base trainning from 768 to 640</li><li>Added support of clothing and R34 version of The Betrayer.<ul><li>I still worked on for absolute flexibility for this model. So, I still updated this model soon, but not to soon as my roadmap of update/making LoRA is mountainous</li><li>Proper R34, unfortunately best works on weight of 0.7 - 0.75, which decreased rate of accuracy of his facial feature. For minimize this, properly tag Valen face features like 'beard, bald, old man', Please note that this thing may not accurate as checkpoint that I use is different with you. In this case I use, OhMenToontastic, Airfuck's Brute Mix, Photon, and Airfuck's Wild Mix.</li></ul></li><li>Simplified Trigger words just to one trigger, the_betrayer</li></ul>

Files with SHA256: f515a2b83277b514036bcc4110facf37cc96743e098026459ca1e179e723c6d7