This project is the result of a custom LoRA training process built entirely from scratch using my own dataset. As you can see from the sample images shared below, the AI-generated characters were created to reflect a highly consistent identity, emotional tone, and stylistic personality.
Every character in this portfolio was generated through a LoRA model I trained myself. The process included organizing a dataset, creating captions, training a model, and integrating it into advanced ComfyUI workflows for both image and video production.
If you have an idea or vision for a character or visual concept that you want to bring to life with AI, feel free to reach out. I’m open to collaborations, creative projects, and custom AI model development.
Contact: LinkedIn Upwork Profile
- ✅ Custom LoRA
- ✅ Image captioning pipeline and dataset preparation
- ✅ ComfyUI workflows for image and video generation
- ✅ Realistic, high-resolution full-body image outputs
- ✅ Short-form video creation
-
Platform: ComfyUI
-
KSampler variations, upscaler
-
Environments:
- Wall background
- Indoor room
- Outdoor street
- Workflow built in ComfyUI
- Duration: 2–4 seconds per video
- Output: 5 .mp4 files
/images: 20 high-quality.jpgimages/videos: 5 short.mp4clips/configs: LoRA training config in JSON/report: Full documentation in PDF and DOCX
