don't use 1.0a it's undertrained, can be useful only for close style, not for character.
Works fine with all sd 1.5 models, train on anything V5 [Prt-RE]
strength 0.8-1, hiresfix or img2img for best results(it's works for every LoRa).
Train models isn't easy, crypto donations on some coffee are welcome:
Boosty, subscribtion with additional content: