I’ve just released a significantly improved version of my personal AI face model — trained using Flux-Dev as the base and fine-tuned with LoRA (Low-Rank Adaptation) for enhanced realism and flexibility in image generation.
Model Link:
What I worked on:
- Used around 30 of my own photos, captured with varied angles, lighting conditions, and facial expressions.
- Created detailed captions for each image to support better LoRA training.
- Each caption file was saved with the same filename as the image, using a .txt extension.
- Trained the model and uploaded it to HuggingFace for secure, accessible storage and sharing.
Gallery of AI-Generated Images
Below is a selection of images generated using the ArunUK V2 model via ComfyUI. These demonstrate the accuracy, expression fidelity, and creative flexibility the model offers.
Interested in creating your own AI avatar?
If you’re looking to:
• Generate AI versions of yourself for creative or professional use
• Build a digital influencer persona
• Explore LoRA-based model training with custom datasets
Feel free to reach out. I’ll guide you through the process or help you build a model from scratch. Access to the ArunUK V2 model is available on request for anyone interested in trying it locally using ComfyUI.
0 Comments