Tensor.Art
Create

FeMix_HassakuXL

CHECKPOINT
Reprint

Updated:

v2.5
v2.4
v2.3
v2.2
v2.1
v2.0
v1.8
v1.7
v1.6
v1.5
v1.4
v1.3
v1.2
v1.1
v1.0
images1
v2.10
v2.9
v2.8
v2.7
v2.6

creator: Koal2

FeMix_HassakuXL: Exploring Blended Styles with the FeMix

FeMix_HassakuXL is a model series built upon base HassakuXL series, pushing creative boundaries with the FeMix blending system. It allows you to generate anime-style images with varying characteristics, all within a single, versatile model family.

FeMix Blending System:

FeMix_HassakuXL introduces different model variants based on the FeMix blending system, offering a spectrum of stylistic nuances. Each version combines two core styles, "A" and "B," in different ratios:

  • Style A: This style provides great lighting and composition. It retains a solid base and is LoRA-Friendly.

  • Style B: Style B is purely unique, it can have many flaws like artifacts. Highres fix or Adetailer is recommended.

Version Breakdown:

  • v1.0: FeMix_HassakuXL_A5B5 (50% A, 50% B): This is a balanced blend, combining the stability and lighting of Style A with the experimental style of Style B. Expect a noticeable influence from both styles, with a moderate risk of artifacts inherited from Style B.

  • v1.1: FeMix_HassakuXL_A7B3 (70% A, 30% B): This blend prioritizes the strengths of Style A, including good LoRA compatibility, while adding a subtle touch of Style B's unique aesthetics. Artifact risk is reduced compared to v1.0.

  • v1.2: FeMix_HassakuXL_A3B7 (30% A, 70% B): This experimental blend leans heavily towards Style B's distinctive style. Be prepared for a higher chance of artifacts and potentially less predictable results with LoRA models. This version is best for exploring unique visuals and accepting potential imperfections.

General Recommendations:

  • Experiment with different prompts and negative prompts to fine-tune the results and control the influence of each style.

  • Use highres fix or Adetailer (or similar tools) to mitigate potential artifacts, especially when using versions with a higher percentage of Style B.

  • The effectiveness of LoRA models may vary depending on the blend ratio. Versions with more Style A will generally provide more predictable LoRA results.

  • Excluded metadata and franchise tags, don't use them. meta data -> something like "highres", franchise tags like "re:zero kara hajimeru isekai seikatsu" are pruned on characters like "rem_(re:zero)".

Recommended Resolutions (SDXL):

  • 1024 x 1024

  • 1152 x 896

  • 896 x 1152

  • 1216 x 832

  • 832 x 1216 (most recommended)

  • 1344 x 768

  • 768 x 1344

  • 1536 x 640

  • 640 x 1536

FeMix_HassakuXL provides a unique way to explore blended styles. Choose the version that aligns with your creative goals and enjoy the journey!

Version Detail

Illustrious
Blend Ratio: 30% Style A, 70% Style B. This blend heavily favors the unique and experimental style of Style B, pushing the model further away from the original Style A. Style B Dominance: Expect a strong expression of Style B's characteristics, including its particular aesthetic, potential for artifacts, and possible composition quirks. Artifact Awareness: Artifacts, especially around the eyes, are more likely with this blend due to Style B's increased influence. Consider using tools like Adetailer aggressively. LoRA Impact Reduced: LoRA models may have a less predictable and potentially weaker effect compared to the A7B3 or A models. LoRA effectiveness is now secondary to capturing the core of Style B. Experimental Focus: This version is primarily aimed at users who want to explore the distinctive style of Style B and are willing to work around potential imperfections. Lighting Effects: Lighting might be less traditionally "cinematic" than in Style A. Expect more unusual and potentially stylized lighting scenarios. Prompt Sensitivity: Prompting may require more experimentation to achieve desired results. Standard prompts that worked well with Style A may need adjustments. In short: Style A3B7 is an experimental blend focused on maximizing the distinctive style of Style B. Be prepared for potential artifacts, a different prompting style, and reduced LoRA control. It's best suited for users who are drawn to Style B's unique look and are willing to experiment.

Project Permissions

Model reprinted from : https://civitai.com/models/1384794

Reprinted models are for communication and learning purposes only, not for commercial use. Original authors can contact us to transfer the models through our Discord channel --- #claim-models.

Related Posts

Describe the image you want to generate, then press Enter to send.