I really missed the aesthetic of the Tunix pony model, and wanted to port it forward to Illustrious, dunno how well it works on most models, as I'm using with Tulpa Toons model. v1 was Trained using my old generations from 2024,v2 I generated a clean dataset in Comfy v1 Lora strength is ?? I used 1.3 with Tulpa Toons for the example image. v2 Lora strength i find even 0.5 is quite strong

Tunix Inspired 3D Style

Generated a much cleaner and more varied dataset, ran for 15 epochs and got a much more useful LoRA

Files with SHA256: 5466c9b184e624cf14a1a9a1f31ab0f1cd67300ab4065a7417e85575546e50b4