piccassol/tetsuo-ai-flux-lora

Tetsuo AI - Flux Image Generation LoRA

Status: Awaiting training images - See issue #1

An uncensored, hyper-realistic LoRA adapter for FLUX.1-dev. Built for high-definition image generation with no content restrictions, aligned with Tetsuo AI's open-source ethos.

Uncensored and Open Source

This model is trained without safety filters or content restrictions. We believe generative AI should be open, transparent, and free from arbitrary censorship. Released under Apache 2.0 for personal and commercial use.

Hyper-Realistic Output

Fine-tuned with a focus on high-definition, photorealistic rendering. Trained in bfloat16 on FLUX.1-dev for maximum detail and visual fidelity. The model produces sharp, cinematic output with realistic lighting, textures, and depth of field.

Character Training

The initial training run focuses on the Tetsuo AI character, a cyberpunk anime female rendered in two modes:

1. High-contrast B&W manga - Heavy ink work, aggressive linework, seinen manga aesthetic with mechanical details
2. 3D cyberpunk portraiture - Volumetric lighting, neon blue highlights, rain, dark metallic surfaces

Trigger word: tetsuo_character

Usage

ComfyUI

1. Download the LoRA .safetensors file
2. Place it in ComfyUI/models/loras/
3. Load Flux dev with the LoRA Loader node
4. Include tetsuo_character in your prompt

Diffusers (Python)

from diffusers import FluxPipeline
import torch

pipe = FluxPipeline.from_pretrained(
    "black-forest-labs/FLUX.1-dev",
    torch_dtype=torch.bfloat16,
)
pipe.load_lora_weights("piccassol/tetsuo-ai-flux-lora")
pipe.to("cuda")

image = pipe(
    "tetsuo_character, cyberpunk portrait, platinum silver hair, neon blue lighting, cinematic, 8k",
    num_inference_steps=20,
).images[0]

Example Prompts

tetsuo_character, cyberpunk portrait, volumetric lighting, neon blue highlights, rain, dark metallic surfaces, black tactical suit, platinum silver hair, pale skin, cinematic, high detail, 8k

tetsuo_character, manga illustration, black and white, heavy ink linework, seinen style, detailed hatching, sharp platinum silver hair, intense eyes, mechanical details

tetsuo_character, walking through cyberpunk cityscape, neon signs, TETSUO CORP glowing in background, rain soaked streets, cinematic composition, ambient orange and blue lighting

Training Details

Parameter	Value
Base Model	FLUX.1-dev (fp8 scaled)
Method	LoRA (Low-Rank Adaptation)
Rank	32
Steps	1500
Learning Rate	1e-4
Optimizer	AdamW
Trigger Word	tetsuo_character
Training Dtype	bfloat16
Bypass Mode	Yes (quantized base)
Framework	ComfyUI built-in training

Links

• GitHub - Training Pipeline
• WAN 2.2 Video LoRA
• Tetsuo AI
• Twitter/X

License

Apache 2.0