This Lora Focuses on Capturing the Low Poly,Shifting Geometric style of Megaman Legends/Rockman DASH In game style Specifically, It uses Captured Frames of a blurry Dazzle recording of the PS1 cut scenes grabbing frames before Dialogue boxes open and HUD is gone.. 942 of them and some ripped textures for in game art/maps.
This Model IS HEAVILY dependent on the Checkpoint for specific subjects IE Animal or Human, Scenery is a constant great job no matter the model i threw at it so far. Also it likes airships.
Detail will be as you expect from the FIRST MML/DASH The second game's Higher Detail models/textures have only 2-5 images in the set thrown in so Temper to how early PS1 looked.. no N64 Port images were used.
As always Temperamental model, Some Loras play nice, Others dont.
Description
A bit more Stable than MML1 Likely due to more varied Minor and major characters fed to it and a lot more environments from the game (Forests/Dungeons/city isn't that great of a spread but just adding Desert and Snow is helping) also more Bot and Airship types are seen, So far I may have mix the Two Training sets for the games for an even more Accurate one. Though I think I may just Boot up my Mister to get Better Images Than using the Dazzle on PS1 hardware. Might look for Codes to disable the HUD for even greater Flexibility, Got plenty of saves to get to important parts of the game.
Definitely will Take longer for Either route. Do what you will with this model and Have fun. This time Training Data should be Available for the model... I Seriously need to Read checkboxes. Apparently if you don't say you own the TD before training it locks it away so its not a toggle Post Training.
Works with other Loras better, Airships are still loved, And due to the res of how I Captured minor Pillar boxes may show up. Highly recommend Against Dazzles for dataset gathering.