Sign In

Midjourney Whisper FLUX LoRA

369
4.0k
10.6k
123
Verified:
SafeTensor
Type
LoRA
Stats
3,981
10,589
Reviews
Published
Aug 17, 2024
Base Model
Flux.1 D
Training
Steps: 1,050
Epochs: 30
Usage Tips
Clip Skip: 2
Strength: 1
Trigger Words
Midjourney_Whisper
A hyper-realistic portrait of a stunning supermodel, featuring a beautifully detailed face with captivating eyes and a serene expression
The image showcases her natural beauty, focusing on her flawless skin and delicate features
The composition blends modern and traditional elements with a neo mode, avant-garde style, characterized by elegant lines, vibrant colors, and exquisite details
Her hair is styled in an intricate and unique fashion, adding to her ethereal and timeless aura
The lighting enhances her features, creating a soft interplay of light and shadow, while the high-resolution quality captures every detail with photorealistic clarity
Hash
AutoV2
498C9E24FC
Created on Civitai
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.

I wanted to bring the elegant atmosphere of Midjourney's beautiful female portraits to Flux, so I decided to experiment by creating a LoRA. I've linked the tags using natural language. For training, I used images that I generated and posted here from Midjourney. Since this is my first attempt, I might have missed some basics, but please feel free to give it a try!

The natural language used for training includes sentences like these.

  • "Midjourney_Whisper, A hyper-realistic portrait of a stunning supermodel, featuring a beautifully detailed face with captivating eyes and a serene expression. The image showcases her natural beauty, focusing on her flawless skin and delicate features. The composition blends modern and traditional elements with a neo mode, avant-garde style, characterized by elegant lines, vibrant colors, and exquisite details. Her hair is styled in an intricate and unique fashion, adding to her ethereal and timeless aura. The lighting enhances her features, creating a soft interplay of light and shadow, while the high-resolution quality captures every detail with photorealistic clarity."

I trained on 30 images this time, and in addition to these descriptions, I added short sentences highlighting each image's beautiful features, such as floral hair accessories, sequins around the eyes, and makeup details.

【example sentence】

  1. Her hair is adorned with beautifully cold-colored white flowers and delicate snowflake crystals, as if dusted like on a Christmas tree, sparkling like jewels. Fine particles of sequins are scattered throughout, enhancing the beauty of these ornaments and making her already stunning face shine even more radiantly.

  2. Her head is beautifully adorned with flowers and plant leaves intricately arranged in a modern and contemporary cool color scheme, serving as a traditional hair accessory. Additionally, vibrant sequins around her eyes make her cute face even more captivating. The patterns on her cheongsam (qipao) embody the essence of ancient Asian traditional art.

  3. Her face is beautifully adorned with a hairpiece combining peacock feathers and plants, featuring a blend of green, purple, and orange hues intertwined with traditional and ceremonial Asian patterns. Similarly, orange eyeshadow shimmers like fine mineral particles around her eyes, enhancing her folkloric feminine beauty. Her irises are exquisitely complex, powerful, and captivating, drawing in anyone who gazes upon them.

I read that 'Danbooru comma separated captioning does not seem to be effective. We’re finding that natural language captioning – narrative style – is a lot more effective for a good output.' Based on this, I avoided using detailed, word-by-word captions for this project.

▼ reference URL "Quickstart Guide to Flux.1"

https://education.civitai.com/quickstart-guide-to-flux-1/

I'm curious to experiment with how natural language can control images, but I feel that in the future, it might be a blend of direct visual instructions and poetic expressions.

【Attention】

Reposting this LoRA on other image generation services without prior written consent is strictly prohibited and may result in legal action. This LoRA is protected by copyright law, and unauthorized use may lead to legal consequences.