Sign In

Melena Maria Rya MoC+

125
1.4k
4
Type
LoRA
Stats
852
Reviews
Published
Jun 23, 2023
Base Model
SD 1.5
Training
Steps: 6,250
Epochs: 1
Trigger Words
MelenaMariaRya
Hash
AutoV2
8FD23DB1BE

‼ IMPORTANT ‼

  • Your voice matters! Show your support and leave a post, comment and most importantly a review! ⭐

  • Custom LORAs available upon request. Just buymeacoffee.☕

  • This is now a + LORA! (details in workflow advice below) 👇(workflow advice below) 👇

📝Biography📝

(generation advice below) 👇

🌟Melena Maria Rya🌟

also known as Maria Ryabushkina, was born on April 16, 1990, in Moscow, Russia. She grew up in a small town near Moscow and had a relatively normal childhood. In her teenage years, she moved to Moscow to study at university, but she dropped out to pursue a career in modeling.

Rya's modeling career took off in 2010 when she was discovered by a photographer on social media. She quickly gained popularity due to her striking looks and exceptional modeling skills. She has modeled for several fashion brands and has appeared in many magazines and photo shoots.

Outside of modeling, Rya enjoys traveling, fitness, and spending time with friends and family. She currently resides in Moscow and continues to work in the modeling industry, where she has established herself as a prominent figure and a source of inspiration for many young models.

💡Workflow Advice💡

🔧Prompt and Sampling🔧

  • What's the + for?

    • Just an easy way to indicate the dataset image for this particular LORA was based off a minimum of 1024px+ quality source images with the LORA trained at 1024,1024 (I may go higher in the future if I can get a properly working configuration).

    • For the most part, you can start with higher base resolutions to capture more details and/or generate higher quality outputs. See some of the gallery examples.

    • You don't have to gen at high resolution, low res generations will be better quality as well.

  • Most images have been generated with LORA weighted @ .75 - 1.1.

  • For photoreal shots, I've found sampling steps best well above 50 give better results.

  • ControlNet is your friend. Embrace the dark side.😈

    • Tile pre-sample and model is good for extreme close ups

    • I use segmentation for lots of my generations now.

🗒️Side Notes🗒️

  • This + series is something new, let me know if I should take the extra time it takes to create these type of LoRAs or just stick with the regular quality.

  • Help me help you: Leave a review if you've used this LORA.

  • Most of my gens are not voodoo or outside of workflows I've indicated here or in my other posts.

    • In other words: Leave a comment if you are having issues getting good output, I'll try to respond with suggestions.