Model Training
Trained custom LoRA styles and operated them as repeatable art-direction systems. Work produced with these models reached 4M+ views on X.
- Role
- Dataset curation, model training, art direction
- Tools
- Krea · ComfyUI · LoRA
- Outcome
- 4M+ aggregate views
Trained custom LoRA styles and operated them as repeatable art-direction systems. Each one locked a visual language and kept the style consistent across a large body of generated work.
Then moved down the stack with ComfyUI, fine-tuning Krea 2 raw and turbo variants at the checkpoint level. Datasets curated, samplers tuned, prompts shaped, all for repeatable output.
