Sign In

Kilian Eng 🎨 Flux.1-D

41

211

4

13

Updated: Apr 21, 2025

stylekilian engflux.1flux1.d

Verified:

SafeTensor

Type

LoRA

Stats

211

4

12

Reviews

Published

Oct 30, 2024

Base Model

Flux.1 D

Training

Steps: 5,800

Usage Tips

Strength: 1

Trigger Words

In style of Kilian Eng

Hash

AutoV2
35744C06A4
seruva19's Avatar

seruva19

The FLUX.1 [dev] Model is licensed by Black Forest Labs. Inc. under the FLUX.1 [dev] Non-Commercial License. Copyright Black Forest Labs. Inc.

IN NO EVENT SHALL BLACK FOREST LABS, INC. BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH USE OF THIS MODEL.

Overview

This F1D LoRA strives to reproduce the artistic style of Kilian Eng, a Swedish visual artist known for his intricate, highly detailed, retro-futuristic sci-fi illustrations. Working under the name 'DW Design', he creates art prints and posters for major clients like HBO, Wired, Warner Bros, and The New York Times. His work blends dreamlike surrealism with physical realism, drawing inspiration from artists like Syd Mead and Moebius. Eng's focus on light and color brings a fantastical, alternative reality to life in his pieces. He regularly releases personal and commissioned works, including prints for video games like Castlevania and Shadow of the Colossus.

Usage

Gallery images were generated with the following settings:

Model: flux1-dev (fp8e4m3fn)
Text Encoder: t5pxxl_fp16
Sampler: euler
Scheduler: 24 steps (normal)
Flux Guidance: 4
LoRA Strength: 1

Training

The LoRA was fine-tuned on an RTX 3090 using the AI Toolkit. A total of 970 images of various resolutions (ranging from 544x544 to 1440x1803) and mostly high quality were used, captioned with MiniCPM-V 2.6 using the prompt "Make a detailed description of this image" with the automatic insertion of "In style of Kilian Eng." at the beginning of each caption.

Training parameters were:

Rank/alpha: 8/8
Optimizer: lion
Steps: 5800
Batch size: 1
Learning rate: 8e-5
Learning rate scheduler: linear
Betas: (0.95, 0.98)
Weight decay: 0.01
Resolution: (512, 768, 1024)