I present my Illustrious character LoRA for Jack Morrison from the Overwatch series.
I made this as a test to see if I could make his hair look softer. It semi-worked but the main differences between my version and other IL/NAI/EPS versions seem to be slightly softer features/color and he came out slightly bulkier due to the NSFW images I included in the training set.
I thought I was done with this LoRA but I was updating other models and decided to modify my training image set for this one to improve hair/face consistency.
V6 is my sixth version:
I modified the training set to also have images NOT tagged as realistic to give better flexibility and remade all nsfw images used for the training.
Recommended LoRA strength: 0.6-0.7
Also, I use an adetailer in my generations for this LoRA.
V5 main tags are as follows:
jackm, 1boy, adult male, mature, very short hairV5 outfit tags are as follows:
scouter, blue pauldrons, blue long coat, blue chestplate, gauntlets, belt, black pants, bootsV5 optional tags:
s76-rifleThe settings I used for my generation testing were 4-5 CFG, 20-24 steps, Euler A - Karras on Forge with an adetailer (face_yolov8s).
I apologize for updating this one so often, the perfectionist in me keeps finding inadequacies.
V5 is my fifth version and I'll give the TLDR version:
I used the same training set as V4 but with additional styled images added to improve 2D/anime generation styles
I added the stubble tag to the training so adding that tag to negative prompt will generally remove his facial hair if you so choose.
Experiment with alternating between tags like anime screenshot, anime coloring, realistic, photorealistic, or no realism tag at all to change up the style.
Recommended LoRA strength: 0.6-0.7
Also, I use an adetailer in my generations for this LoRA.
V5 main tags are as follows:
jackm, 1boy, adult male, mature, short hairV5 outfit tags are as follows:
scouter, blue pauldrons, blue long coat, blue chestplate, gauntlets, belt, black pants, bootsV5 optional tags:
s76-rifleThe settings I used for my generation testing were 3-4 CFG, 20 steps, DPM++ 2M SDE - Karras on Forge with an adetailer (face_yolov8s).
V4 is my fourth version and attempt to even further refine the training set and overall quality. My goal was to get a nice face and hair consistency and finally succeeded on getting a VERY nice consistency in quality with his NSFW generations with this model so I'm very happy with this one.
I completely removed all training tags for hair/eye color, muscular, body hair, and other such tags so his appearance would be more baked in with fewer words (outfit is unchanged).
Experiment with alternating between realistic, photorealistic, or no realism tag at all as I tried to strike a balance.
Recommended LoRA strength: 0.6-0.7
Also, I use an adetailer in my generations for this LoRA.
V4 main tags are as follows:
jackm,1boy,adult maleV4 outfit tags are as follows:
scouter, blue pauldrons, blue long coat, blue chestplate, gauntlets, belt, black pants, bootsV4 optional tags:
s76-rifleThe settings I used for my generation testing were 3-4 CFG, 20 steps, DPM++ 2M SDE - Karras on Forge with an adetailer (face_yolov8s).
V3 is my third version and is my attempt to further refine his face. I completely revised my NSFW training images and I included many more bare face (non-AI) images to drastically improve the accuracy of the face. You no longer need the stubble tag to actually get his subtle facial stubble.
Recommended LoRA strength: 0.7
Also, I use an adetailer in my generations for this LoRA.
V3 main tags are as follows:
jackm,1boy,mature male, blonde hair, very short hair, blue eyesV3 outfit tags are as follows:
scouter, blue pauldrons, blue long coat, blue chestplate, gauntlets, belt, black pants, bootsV3 optional tags:
rifleThe settings I used for my generation testing were 3-4 CFG, 20 steps, DPM++ 2M SDE - Karras on Forge with an adetailer (face_yolov8s).
V2 is my second version and produces great results. I refined the NSFW training images for this one because I wanted his face to be more consistent and look slightly more mature. The hair looks slightly more defined and the stubble more obvious. I did not add any other outfits but I did manage to slim him down so feel free to add bara or muscular tags to make him bulkier.
Recommended LoRA strength: 0.7
Also, I use an adetailer in my generations for this LoRA.
V2 main tags are as follows:
jackm,1boy,mature male, blonde hair, very short hair, blue eyes, stubbleV2 outfit tags are as follows:
scouter, blue pauldrons, blue long coat, blue chestplate, gauntlets, belt, black pants, bootsV2 optional tags:
rifleThe settings I used for my generation testing were 3-5 CFG, 20 steps, DPM++ 2M SDE - Karras on Forge with an adetailer (face_yolov8s).
V1 is my initial version and produces great results. I only trained his standard outfit so if you want additional outfits and versions I recommend you try this model by @Queria or this model by @starkomi20. I tried to train his rifle but don't expect much with that tag. 😜
Recommended LoRA strength: 0.7
Also, I use an adetailer in my generations for this LoRA.
V1 main tags are as follows:
jackm,1boy,mature male, blonde hair, very short hair, blue eyes, stubbleoptional - add "muscular" and "bara" to negative prompt if he looks too bulky (especially in NSFW or shirtless shots).
V1 outfit tags are as follows:
scouter, blue pauldrons, blue long coat, blue chestplate, gauntlets, belt, black pants, bootsV1 optional tags:
rifleThe settings I used for my generation testing were 3-5 CFG, 20 steps, DPM++ 2M SDE - Karras on Forge with an adetailer (face_yolov8s).
Description
V1 is my initial version and produces great results. I only trained his standard outfit so if you want additional outfits and versions I recommend you try this model by @Queria or this model by @starkomi20. I tried to train his rifle but don't expect much with that tag. 😜
Recommended LoRA strength: 0.7
Also, I use an adetailer in my generations for this LoRA.
FAQ
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.