💰Oligarchy💰

Version: 1.0 – Released 2025-01-20

Oligarchy is my latest photorealistic Frankenstein of a merge, designed to crank out ultra-sharp, realistic images. I took my custom LoRAs (some trained from scratch on my own data) and meshed them with a carefully curated set of open-source LoRAs. The result? A model that dishes out lifelike images for portraits, street photography, scenic shots—whatever you can dream up. This model is all about capturing the absolute best of what I think a photoreal merge should have.

Under the Hood

• Base Model: Flux Dev (foundation)

• LoRA Stack: 50 LoRAs total

  • 15 hand-trained by yours truly

  • 35 borrowed from open-source, because sharing is caring

• Resolution Sweet Spot: 840×1232 (though it can stretch beyond that if you have the juice)

Merge Approach

1. Core Merge: Flux Dev + my personal LoRAs, especially the ones focused on photoreal textures (skin, fabric, hair, lighting).

2. Open-Source Goodies: Mixed in specialized environment and detail-focused LoRAs for max immersion.

3. Final Tuning: Balanced everything so you won’t get weird color shifts or style conflicts.

Highlights

• Hyper-Real: Crisp details on faces, backgrounds, and everything in between.

• Stability: Fewer random artifacts, more consistent composition.

• Prompt Flexibility: Works for moody portraits, bright landscape shots, or even close-up product renders.

• Plays Nice with Upscaling: Compatible with your favorite upscalers and in-painting workflows.

Shout-Outs

• Flux Dev Creators: Couldn’t do this without that awesome base.

• LoRA Contributors: Thank you for your open-source wizardry.

• Testers & Supporters: You know who you are—your feedback is gold.

Getting Started

1. Grab the Model: Throw Oligarchy.gguf (or .safetensors) into your comfy-ui models/unet folder.

2. Prompt Tips:

   • Example prompt:

     A cinematic portrait of [subject], ultra-detailed, photoreal, shot on 50mm lens

   • Tweak your weights or add style LoRAs to get the exact look you want.

3. Sampler/schedulers: Euler/simple or uni_pc_bh2/beta are my personal go-tos, but experiment away.

Version Notes

• v1.0 (2025-01-20):

  • Initial public release.

  • Thorough testing to ensure stable color, minimal artifacts, and that “so real you can feel it” vibe.

Struggling with flux prompts? try my flux prompt craft GPT bot

Do you use the GGUF? Try these workflows. GGUF T5 CLIP only | GGUF UNET+GGUF T5

Looking for a Latent node that's optimized for FLUX, SD3 & SDXL? Bob's FLUX Latent Optimizer Node

For optimal results, I recommend using 20-32 steps.

Remember to also download the custom CLIP L! The current version of CLIP L is V5

Want an alternative CLIP L? Try this one by Zer0int