Canserbero's image and legacy are not used for profit.
This fine-tuning was trained using 60 dataset images on an L40 at Massed Compute.
Trigger word: Canserbero.
Example Prompt: Canserbero. A young latin man with an oval-shaped face, soft but defined jawline, and medium skin tone with a warm undertone. He has intense, almond-shaped olive green eyes with golden undertones that reflect depth and emotion, framed by naturally thick, straight eyebrows. His nose is straight with a slightly rounded tip, proportional to his face. His lips are medium-sized, with a thinner upper lip and a fuller lower lip. He has a subtle mustache and goatee, neatly groomed, adding character to his face. His hair is dark brown to black, medium-length, wavy or slightly curly, falling naturally to the sides. + [YOUR PROMPT].
Description
Comments (5)
Nice tribute bro.
I'm curious, is this not possible to achieve with a LoRA? What's the reason for a full fine-tune for a character concept? Better quality than what can be achieved with a LoRA?
Fine-tuning the full FLUX Dev model gives much better results than LoRA, especially for character concepts. LoRA tends to overfit, meaning it struggles to create variations beyond the training images. A fine-tuned model, on the other hand, can adapt better to new prompts, expressions, and artistic styles.
@JotaTerrasa Cool! Thanks for clarifying. :)
Awesome job Pablo! Excellent execution of Finetuning!
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.