I made a Matt (Aqua Admin Matt) of Team Aqua LoRA from Pokemon. He's known for his muscular body, his personality, and the way he speak (especially to Archie). I made this beefcake dude wth 130+ images with mostly from Google Images and some various image sharing sites. Mostly trained from fanart. Matt also known as bara material.
Trigger Words (choose one)
aqua_admin_matt, matt \(pokemon\)Settings I used:
Weight: 0.7 - 0.9
Sampling Method: DPM++ 2M SDE Karras
Sampling steps: 22 - 25
Resolution: 768x768 or 960x960 (768x960)
CFG Scale: 7
Checkpoints
Mature Male Mix
Airfuck's Brute Mix
AnyLoRA + Pokemon Style LoRA
Description
I updated this model by increasing image quantity back and removal some bad images. This update is minor update. Here's some update notes:
Increase image used in dataset from 120+ to 130+ with some removal bad images and addition of new images with various poses and styles
Increase steps in training
Simplified trigger word just into 1 activation tag
v1.1:
I updated this model. I noticed that in my dataset there's some incredibly bad image that affects the quality of dataset, hence the decrease number of dataset. Despite decreased image quantity, there also image addition that has much higher quality and all images has bein resized. Here's some breakdown of this model update:
Increased image resolution up to 4x from original
Decreased Dataset from 130+ to 120+ images
There's still additional new images that has high quality
Increased LoRA size 2x
Resolution Train Image Increased to 640
Simplified namefile from Aqua_Admin_Matt-15 to Matt-15
v1.0:
130+ images
Google Images
Pinterest
Pixiv
Tumblr
Epoch: 15
Steps: 10
FAQ
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.