Sign In

Nepotism

691
12.1k
284
Verified:
SafeTensor
Type
Checkpoint Merge
Stats
411
Reviews
Published
Jan 25, 2025
Base Model
Flux.1 D
Training
Steps: 69
Epochs: 420
Hash
AutoV2
C3A24F33C4
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.

Please post images to the model page; all buzz earned goes towards training LORA for the community

!!GGUF file download guide is in the version notes of the GGUF version!!

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 with Nepotism, I recommend:

What's New in XI?

Nepotism XI makes X look like a beta test. I couldn’t leave X half-baked, so I fixed it. Now you get:

  • Higher Fidelity: Everything looks cleaner, sharper, and just overall better. Textures pop more, lighting feels more natural, and intricate elements are handled with precision. Whether it's photorealism or stylized work, XI delivers top-tier visuals without breaking a sweat.

  • 3x Style Range: XI expands its creative scope by offering a massive increase in style adaptability. It handles photorealism, anime, cartoons, and abstract art seamlessly, giving you way more freedom to create without limitations.

  • Finer Detail Focus: No more weird artifacts or random noise—just crisp, clean outputs that maintain detail even in complex compositions. You’ll notice sharper edges, better color depth, and more lifelike renderings.

  • Balanced Aesthetic: Unlike X, XI doesn’t force anime when you don’t want it—but it might still throw it in if it thinks the prompt leans that way. It’s better at balancing realism and style, but it’s not psychic, so be clear with what you want.

  • Better Prompt Adherence: Say goodbye to frustration—XI understands complex and abstract prompts better than ever. Whether you’re aiming for subtle artistic nuances or specific stylistic choices, the model actually gets what you mean.

  • Improved Speed: Faster processing without sacrificing quality. Even complex prompts render quickly, making XI more efficient for rapid iterations and professional workflows.

Why Nepotism Fux Stands Out:

- Balanced Precision: This merge uses FP8, producing images that closely resemble FP16 quality at a fraction of the time. Perfect for users with mid-range PCs who want Flux1Dev-level results without the resource drain.

- Efficiency: At 20 steps, generate high-quality images in just 16-22 seconds on a 4080 GPU, compared to the 80-150 seconds typical with Flux1Dev FP16.

X Recap:

  • Moved away from its strong photorealism bias and embraced anime, cartoons, and 3D modeling.

  • Improved balance across styles, offering more creative flexibility.

  • Prompt adherence saw improvements but still had occasional biases.

  • Versatile but sometimes leaned into anime unexpectedly.

  • Solid upgrade over previous version

V9 Recap:

  • Enhanced Photo-Realism: Refines every detail, from textures to lighting, ensuring your creations look lifelike and captivating.

  • Improved LoRA Compatibility: Retains its sensitivity to LoRA weights, allowing for greater precision when fine-tuning outputs.

  • Better Noise Reduction and Detail Focus: Generates outputs with less noise and a sharper focus on fine details

  • Improved Composition and Style Adherence: More cohesive compositions and adherence to specific artistic styles

  • Expanded Creative Versatility: Exceptional results for abstract concepts, anime, and cartoons

  • Performance:

    • Cold Load (No LoRAs): 1.05-1.11s/it

    • Cold Load (With LoRAs): ~2.10-3.45s/it, improving to 1.05-1.35s/it post-load.

V8 Recap:

  • NSFW Content Removed: The model is now entirely focused on safe-for-work content.

  • Mild Photorealism Bias: V8 leans towards producing photorealistic images.

  • Increased Sensitivity to LoRAs: More responsive to LoRA weights, allowing you to achieve great results with smaller adjustments.

  • Read the full update article here.

  • Performance:

    • Cold Load (No LoRAs): 1.05-1.11s/it

    • Cold Load (With LoRAs): ~2.10-3.45s/it, improving to 1.05-1.35s/it post-load.

V7 Recap:

  • Huge TY to @jurdn for helping me with the V4-V7 Q8/Q4KS GGUF's!

  • LORA Compatibility Improvements: Enhanced efficiency and versatility with LoRAs.

  • NSFW Fine-Tuning: Further refinement of NSFW outputs without compromising on quality.

  • Improved Photorealism & Detailing: Finer details and sharper outputs.

  • Performance:

    • Cold Load (No LoRAs): 1.05-1.11s/it

    • Cold Load (With LoRAs): ~2.10-3.45s/it, improving to 1.05-1.35s/it post-load.

V6 Recap

  • LORA Compatibility Fixed: Addressed LORA-related issues, achieving full compatibility.

  • NSFW Improvements: Continued efforts to enhance NSFW content without compromising SFW content.

  • Detail and Accuracy: Further sharpened details, significantly better handling of hands, and enhanced image quality.

  • Speed and Performance: Improved speed and stability across various workflows, especially with LORAs.

  • Performance:

    • Cold Load (No LORAs): 1.00-1.25s/it.

    • Cold Load (With LORAs): ~3.15-5.30s/it, improving to 1.10-1.45s/it.

V5 Recap

  • Sharper fine details and an overall boost in quality.

  • Hands improved significantly—no moreLESS weird fingers!

  • Performance:

    • Cold Load (No LORAs): 1.07-1.25s/it.

    • Cold Load (With LORAs): ~3.25-5.45s/it, improving to 1.13-1.56s/it.

V4 Recap

  • Improved accuracy and stylization for better prompt-to-output consistency.

  • NSFW content refined (still no XXX).

  • LORA support improved.

  • Performance highlights:

    • Cold Load (No LORAs): 1.00-1.05s/it.

    • Cold Load (With LORAs): ~3.25-5.45s/it.

V3 Recap:

  • Custom LORAs and CLIP L from NepotismXL.

  • Recommended steps: 20-32 for optimal results.

  • Full Version: Consistent, high-quality outputs.

  • Pruned Version: More NSFW but also more artifacts and deformities.

Key Advantages:

  • Balanced Precision: FP8 offers images close to FP16 quality with reduced time.

  • Efficiency: High-quality images in 25-45 seconds on a 4080 GPU, versus 80-150 seconds using Flux1Dev FP16.

V2 Recap:

- A merge of 71% Flux.1 Dev and 29% Flux.1 Schnell, further refined with an enhanced CLIP.

- AIO Version: Combines NepotismFUX V2 DiT, ae vae, t5xxl, and NepotismXL V2 CLIP L into a single, all-inclusive safetensor file.

V1 Recap:

- The original NepotismXL CLIP L for SDXL, which paved the way for the enhanced capabilities in V2 and V3.

How to Get FUX'D Up:

- AIO Version:

- STEP 1: [CLICK HERE FOR THE FUX-ING WORKFLOW]

- STEP 2: PROFIT

- DiT & CLIP Separately:

- STEP 1: [CLICK HERE FOR THE FUX-ING WORKFLOW]

- STEP 2: PROFIT

Why 71% Dev & 29% Schnell?

- Enhanced Speed with Quality: The 71% Dev/29% Schnell merge delivers high-quality images at faster speeds, making it perfect for rapid prototyping and iterative design.

- Resource Efficiency: FP8 precision allows users with lower-end hardware to generate top-tier images without sacrificing quality.

- Scalability: This model adapts well to various hardware configurations, broadening its accessibility.

Considerations:

- Image Quality: While FP8 may reduce some image quality compared to FP16, the merge maintains a significant portion of Dev's high-quality output.

- Prompt Adherence: Strong prompt adherence is retained from the Dev model while benefiting from Schnell's speed and efficiency.

Solution:

- Image Quality & Prompt Adherence: By integrating a highly precise CLIP L from SDXL, the model achieves compositionally similar results to Dev at FP16, especially when using CLIP L with Dev FP16.

Tested & Running on:

- Nvidia 4090 (24GB VRAM) 64GB of RAM

- Nvidia 4080 (16GB VRAM) 32GB of RAM

- Nvidia 3080 TI (12GB VRAM) 32GB RAM

Note: This model isn't designed for XXX content (reliably anyways), though it can produce R and X-rated images.