I present my Illustrious character LoRA for Emmrich Volkarin from Dragon Age: The Veilguard.<hr /><hr />V1 is my initial version and produces decent results. The red glove can be a bit random at times...but there's only so much that can be done about that (asymmetry does not play nice with AI).Recommended LoRA strength: 0.7 Also, I use an adetailer in my generations for this LoRA.V1 main tags are as follows:<pre><code>e_volkarin, 1boy, mature male, hazel eyes, mustache, grey hair, black hair, multicolored hair, very short hair</code></pre>optional - add "muscular" to negative prompt if he looks too bulky (especially in NSFW or shirtless shots).V1 outfit tags are as follows:<pre><code>green vest, yellow shirt, bracelets, multiple rings, single red glove, red waist sash, tan pants, red boots</code></pre>V1 optional tags:<pre><code>mage staff, garden, cemetery</code></pre>The settings I used for my generation testing were 3-5 CFG, 20 steps, DPM++ 2M SDE - Karras on Forge with an adetailer (face_yolov8s).Hide

Dragon Age: The Veilguard - Emmrich Volkarin - Illustrious

V1 is my initial version and produces decent results. The red glove can be a bit random at times...but there's only so much that can be done about that (asymmetry does not play nice with AI).Recommended LoRA strength: 0.7 Also, I use an adetailer in my generations for this LoRA.

DAV - Emmrich Volkarin - Illustrious_1-000006.safetensors

Emmrich 4.zip

These are my rough attempts at a character LoRA for Emmrich Volkarin from Dragon Age: The Veilguard.<hr />V3 I tried to further refine the training tags, being more selective with training images, and trying to improve the background options by including scenery shots in the training.<ul><li>I tried training for hazel eyes (instead of brown), as I noticed I sometimes see green in-game, so I wanted it to be more accurate.</li><li>I tried to training for both gray AND black hair (as well as the multicolored hair tag) since his hair definitely still has patches of black/dark gray so i felt just having a gray hair tag wasn't accurate.</li><li>The likeness is consistent and the bracelets/glove are still random at times but not too terrible.</li><li>The mage staff...is a nightmare to work with on the more realistic checkpoints (like homosimile) but it seems "okay" on 2D checkpoints (like AutismMix and the Pony V6 checkpoints).</li><li></li></ul>I recommend keeping the LoRA strength around 0.7-0.8 and CFG between 4-7.An adetailer is also recommended for the more zoomed out shots.V3 main tags are as follows:<pre><code>e_volkarin, mature male, hazel eyes, mustache, grey hair, black hair, multicolored hair, very short hair</code></pre>optional - add "muscular" to negative prompt if he looks too bulky (especially in NSFW or shirtless shots).V3 outfit tags are as follows:<pre><code>green vest, yellow shirt, bracelets, multiple rings, single red glove, red waist sash, tan pants, red boots</code></pre>V3 optional tags:<pre><code>staff, garden, cemetery</code></pre>The settings I used for my V1 generation testing were 4-7 CFG, 20 steps, Euler A - Automatic using Forge.<hr /><hr />V2 (clip skip 1) I tried adjusting the training tags and also added more images to the training to make the LoRA more versatile when used with 2D and 2.5D checkpoints. I recommend keeping the LoRA strength around 0.8 and CFG between 3-5.An adetailer is also recommended for the more zoomed out shots.V2 main tags are as follows:<pre><code>e_volkarin, mature male, hazel eyes, mustache, grey hair, short hair</code></pre>optional - add "muscular" to negative prompt if he looks too bulky (especially in NSFW or shirtless shots).V2 outfit tags are as follows:<pre><code>green vest, yellow shirt, bracelets, multiple rings, single red glove, red waist sash, tan pants, red boots</code></pre>V2 optional tag<pre><code>staff (weapon)</code></pre>The settings I used for my V1 generation testing were 3-5 CFG, 20 steps, DPM++ 2M - Karras using Forge.<hr /><hr />V1 (clip skip 1) was my first attempt for this LoRA and generates decent results but not necessarily the greatest (the mustache sometimes comes out thicker than it should be). I recommend keeping the LoRA strength around 0.8 and CFG between 5-7.An adetailer is recommended and you may also need to add "fingerless gloves" to negative prompts and "closed mouth" to positive prompts.V1 main tags are as follows:<ul><li>e_volkarin - (subject tag)</li><li>mature male</li><li>hazel eyes</li><li>mustache</li><li>grey hair</li><li>short hair</li><li>optional tag - "closed mouth" - you may need this if he keeps generating with an open mouth as if he's speaking.</li></ul>V1 outfit tags are as follows:<ul><li>green vest</li><li>yellow shirt</li><li>collar chain (jewelry)</li><li>gloves - (you may need to add "fingerless gloves" to negative prompt if the red glove keeps rendering oddly)</li><li>red waist sash</li><li>tan pants</li><li>red boots</li><li>optional tag - "staff"</li></ul>The settings I used for my V1 generation testing were 5-7 CFG, 30 steps, DPM++ 2M - Karras using Forge.

Dragon Age: The Veilguard - Emmrich Volkarin - Pony

V3 I tried to further refine the training tags, being more selective with training images, and trying to improve the background options by including scenery shots in the training. <ul><li>I tried training for hazel eyes (instead of brown), as I noticed I sometimes see green in-game, so I wanted it to be more accurate.</li><li>I tried to training for both gray AND black hair (as well as the multicolored hair tag) since his hair definitely still has patches of black/dark gray so i felt just having a gray hair tag wasn't accurate.</li><li>The likeness is consistent and the bracelets/glove are still random at times but not too terrible. </li><li>The mage staff...is a nightmare to work with on the more realistic checkpoints (like homosimile) but it seems "okay" on 2D checkpoints (like AutismMix and the Pony V6 checkpoints).</li></ul>I recommend keeping the LoRA strength around 0.7-0.8 and CFG between 4-7.An adetailer is also recommended for the more zoomed out shots.

Files with SHA256: 741a3c0aed3bf43c99c30bccabbec11a800a6a1ac2973b0db11cc5bef6902aad