Sign In

FolderDesign

18
176
12
7
Updated: Sep 9, 2024
conceptuxuidesign
Verified:
SafeTensor
Type
LoRA
Stats
118
11
77
Reviews
Published
Aug 17, 2024
Base Model
Flux.1 D
Training
Steps: 4,000
Trigger Words
Folder icon
Hash
AutoV2
A924EEFFC6
default creator card background decoration
jerel
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.

This LoRA is my first attempt at a LoRA of Windows Folder Design.

Civitai's default settings made this model very small and made it not keep the consistency of its dataset. In about 8 hours, there should be a better, more consistent model up.
The dataset consists of 215 images of consistent Windows folder designs captioned by gpt4-o using sd-caption-helper (https://sd-caption-helper.vercel.app/).

The captions are full sentences to match flux. But I would imagine that "folder icon" would help as a trigger word.

Taking suggestions for LoRAs for SD 1.5, SDXL and Flux. If you have something, a style, a character or a concept that could be interesting to make a LoRA of please give me ideas.

ComfyUI workflow: https://openart.ai/workflows/ailab/flux-lora-simple/gErKNyjsnetZpEgaMTrV

Changelog:

Version 1.5:
Improved consistency through higher rank size

Workflow

1. From over 2.000 images, I've picked images I thought would enhance the LoRA the most.
2. I've used SD-caption-helper to caption the images.
This is the prompt I've used:

Write detailed full sentence caption for this image while keeping it as short as possible avoiding all unnecessary words like "the". Mention the Colors, design, logo company name and mention all text depicted in the image in quotes. The caption should be between 40 and 75 words long. Don't mention the "image" just get straight into the description like "A ...". Avoid overly flowery language. Give me the description only. No unnecessary mention of you task. Keep it plain and simple and short. Don't wrap the whole prompt in any quotes or similar. Only describe what can be seen and not what can be assumed unless its sure. Make full sentences as simple as possible.


3. I've used Civitai Flux training for version 1.0. ai-toolkit for version 1.5

PS: If you want the dataset please contact me. I just don't want to get CivitAI in copyright trouble.