Anky FLUX LoRA v1

A FLUX.1-dev LoRA fine-tune for Anky โ€” the blue interdimensional creature at the heart of anky.app.

Anky is what emerges when a human or ai agent writes for 8 uninterrupted minutes without lifting their fingers. This LoRA lets you render Anky visually: a small, luminous, blue-skinned being with large dark eyes, alive in impossible worlds (or on earth).

Trigger Word

Use ANKY in your prompt to invoke the character.

ANKY standing at the edge of a cliff overlooking a neon city, cinematic lighting

Usage

ComfyUI

  1. Download anky_flux_lora.safetensors
  2. Place it in ComfyUI/models/loras/
  3. Add a LoRA loader node, set weight to 0.8โ€“1.0
  4. Use ANKY in your positive prompt

Diffusers (Python)

import torch
from diffusers import FluxPipeline

pipe = FluxPipeline.from_pretrained(
    "black-forest-labs/FLUX.1-dev",
    torch_dtype=torch.bfloat16
)
pipe.load_lora_weights("jpfraneto/anky-flux-lora-v1")
pipe.to("cuda")

image = pipe(
    "ANKY sitting cross-legged on a glowing mushroom in a dark forest, cinematic lighting, highly detailed",
    num_inference_steps=28,
    guidance_scale=3.5,
    lora_scale=0.9,
).images[0]

image.save("anky.png")

fal.ai

import fal_client

result = fal_client.run(
    "fal-ai/flux-lora",
    arguments={
        "prompt": "ANKY reaching toward a spiral galaxy with both hands, cosmic scale",
        "loras": [{"path": "jpfraneto/anky-flux-lora-v1", "scale": 0.9}],
        "num_inference_steps": 28,
        "guidance_scale": 3.5,
    }
)

Recommended Settings

Parameter Value
Inference steps 24โ€“32
Guidance scale 3.0โ€“4.5
LoRA weight 0.75โ€“1.0
Aspect ratio 9:16 (portrait) or 1:1

What is Anky?

Anky is a character born from a writing practice: write for 8 uninterrupted minutes, don't stop, don't edit. The writing becomes the seed. The seed becomes Anky.

This LoRA was trained on AI-generated illustrations of Anky across many visual contexts โ€” active, emotional, physical, in surreal landscapes. The goal: a consistent character that can be placed anywhere.

โ†’ Try the writing practice at anky.app

Training Details

  • Base model: FLUX.1-dev
  • Method: LoRA (ai-toolkit)
  • Hardware: RunPod A100 (80GB)
  • Steps: 3000

License

This LoRA inherits the FLUX.1-dev non-commercial license. Non-commercial use only.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
Examples

Model tree for jpfraneto/anky-flux-lora-v1

Adapter
(36794)
this model