<h1 id="azoomer-style-is-a-lora-model-for-mimicking-azoomer's-art-style-of-furry-characters.">Azoomer Style is a LoRA model for mimicking Azoomer's art style of furry characters.</h1><h2 id="noobai-v1">NoobAI V1</h2>Main Trigger Word: AzoomerSecondary Trigger Words: Monochrome, Flat RenderAzoomer has many sketches which are flat/monochrome. Use the secondary trigger word in positive prompts to activate this alternate style or place it in negative prompt to exclude it completely.Recommended LoRA Parameters for YiffyMixV62:SDXL Model: YiffyMix V62 VAE: Integrated into base model Clip Skip: Not needed LORA Weight for Azoomer SDXL V1: 0.5-0.6 for mixing, 0.7+ for full effect, 1.0+ not recommended LORA Weight for SeaArt Quality Tags : 0.5 (Not necessary) LORA Weight for NAI Stabiliser: 0.5 (Not necessary) CFG: 5 Recommended YiffyMix V52 Parameters Res: 1024x1024, 1280x1280, (1:1), 1024x1536 (2:3) Sampler: Euler A Schedule type: SGM Uniform HiRes Fix: 2-4x <a target="_blank" rel="ugc" href="https://civarchive.com/models/147806?modelVersionId=164882">UltraMix_Smooth</a>, 0.3 Denoise, 15 HiRes Steps, 5 HiRes CFG, Post Process Upscale: 2-4x <a target="_blank" rel="ugc" href="https://civarchive.com/models/147806?modelVersionId=164882">UltraMix_Smooth</a> Negative Embeddings: <a target="_blank" rel="ugc" href="https://civarchive.com/models/16807?modelVersionId=53718">worst quality</a><h2 id="sdxl-v1">SDXL v1</h2>Recommended LoRA Parameters for YiffyMixV52 (SDXL):SDXL Model: YiffyMix V52 VAE: Integrated into base model Clip Skip: 1 LORA Weight for Azoomer SDXL V1: 0.7 - 0.8 LORA Weight for SeaArt Quality Tags : 0.5 CFG: 5 Recommended YiffyMix V52 Parameters Res: 1024x1024 (1:1), 1024x1536 (2:3) Sampler: Euler Schedule type: SGM Uniform HiRes Fix: 2-4x <a target="_blank" rel="ugc" href="https://civarchive.com/models/147806?modelVersionId=164882">UltraMix_Smooth</a>, 0.3 Denoise, 15 HiRes Steps, 5 HiRes CFG, Post Process Upscale: 2-4x <a target="_blank" rel="ugc" href="https://civarchive.com/models/147806?modelVersionId=164882">UltraMix_Smooth</a> Negative Embeddings: <a target="_blank" rel="ugc" href="https://civarchive.com/models/16807?modelVersionId=53718">Check SeaArt Quality Tags</a><h2 id="pdxl-v1">PDXL v1</h2>Recommended LoRA Parameters for PonyDiffusion V6 (SDXL):SDXL Model: Pony Diffusion V6 VAE: SDXL.vae Clip Skip: 1-2? (Ive always used 1 and it works fine despite known issues with PDXL) LORA Weight: 0.7 - 0.8 (0.8+ looks fried) Recommended Pony Diffusion V6 Parameters Res: 1024x1024 (1:1), 1024x1536 (2:3) HiRes Fix: 1-2x ESRGAN 4x, 0.4 Denoise Strength Post Process Upscale: 2-4x <a target="_blank" rel="ugc" href="https://civarchive.com/models/147806?modelVersionId=164882">UltraMix_Smooth</a> Negative Embeddings: <a target="_blank" rel="ugc" href="https://civarchive.com/models/16807?modelVersionId=53718">Not needed</a>PROMPT SYNTAX (IMPORTANT)Put these phrases at the start of the prompt to ensure desirable image generation with the LoRA.<pre><code>azoomer, score_9, score_8_up, score_7_up, score_6_up,</code></pre>LoRA was trained on E621 prompts so use that tagging convention in the prompt for best images.<h2 id="sd-1.5-v4">SD 1.5 V4</h2>Recommended LoRA Parameters for YiffyMix V44 (SD 1.5): 1.5 SD Model: YiffyMix V44 Clip Skip: 1 LORA Weight: 0.8 - 1 Recommended YiffyMix V44 Parameters Res: 768x768 (1:1), 768x1152 (2:3) HiRes Fix: 1-2x ESRGAN 4x, 0.4 Denoise Strength Post Process Upscale: 2-4x <a target="_blank" rel="ugc" href="https://civarchive.com/models/147806?modelVersionId=164882">UltraMix_Smooth</a> Negative Embeddings: <a target="_blank" rel="ugc" href="https://civarchive.com/models/16807?modelVersionId=53718">DeformityV6</a>, <a target="_blank" rel="ugc" href="https://civarchive.com/models/187118?modelVersionId=210124">bwu</a>, <a target="_blank" rel="ugc" href="https://civarchive.com/models/7808?modelVersionId=9208">easy negative</a> SubStyles: The LORA will produce a mixture of greyscale and colourful images by default as the training data includes these two. To force a specific style such as sketches, use the additional prompts "Monochrome" and "Simple Background" and "Simple shading".To force colourful images such as those found in the gallery, use the additional prompts "colorful" and "high contrast". However, it is to note that if you original prompt includes colour like "purple body" or "yellow eyes" the LORA will default to being colourful anyways.Limitations: <s>This LORA may only accurately reproduce the art style of Azoomer based on the types of characters Azoomer has drawn</s>. V4 has somewhat mitigated this. Some characters and species may still be incompatible. Check the LORA metadata for the tokens of the training dataset. Any LORA that tries to add detail may cause adverse effects such as "Add Detail LORA" Version History: V2 - 74 Images (30 Steps, 5 Epochs), YiffyMix V43 V3 - 188 Images (30 Steps, 5 Epochs), YiffyMix V44 V4 - 188 Images (30 Steps, 5 Epochs), YiffyMix V44 (FIXED UNWANTED TEXT / WATERMARKS)PDXL V0.5 - 48 Images (30 Steps, 10 Epochs), Pony Diffusion V6 (TEST LoRA) PDXL V1 - 188 Images (30 Steps, 6 Epochs), Pony Diffusion V6SDXL V1 - 188 Images (20 Steps, 10 Epochs), YiffyMixV52NoobAI V1 - 296 Images (1 Step, 7 Epochs), YiffyMix62

Azoomer Style LoRA | NoobAI, PonyXL, SDXL, SD 1.5 |

Fixed Text / Watermark Issues<ul><li>Watermarks should not appear in 95% of the images generated. </li><li>No need to put watermark / text in negative prompt</li><li>Removal of watermarks more effective than V3.0 and V2.0</li></ul>

Azoomer V4.safetensors

SD 1.5 v4.0

NoobAI V1

SDXL V1.0

PDXL V1

PDXL v0.5

SD 1.5 v3.0

SD 1.5 v2.0

creator: FatKiwi817A LORA model for reproducing the artist Azoomer's style for furry characters. Recommended Usage: Base SD Model: YiffyMix V43 (Was the base model the LORA was trained on) Clip Skip: 1 LORA Weight: 0.8 (Increase weight if your prompts are more than 100 tokens long) Use recommended YiffyMix V43 T2I Settings such as Res: 768x768 Clip Skip 1 Negative Embeddings: DeformityV6, bwu, easy negative Substyles: The LORA will produce very colourful images when using minimal prompt tokens. If you want to mimic the exact style of Azoomer (which contains grayscale images), put "monochrome" in with the promptLimitations: This LORA will only accurately reproduce the art style of Azoomer based on the types of characters Azoomer has drawn. Some characters may be incompatible. Check the LORA metadata for the tokens of the training dataset. In testing, the LORA is not compatible with the Add Detail LORA as both of these models are polar opposites on what they try to achieve. As prompt complexity increases, the strength of the LORA way fall behind, to combat this, increase the LORA weight to 0.9 or even 1. For most images, leaving it on 0.8 or 0.7 produces great results. Further revisions may come as I increase the dataset sample size.

Azoomer Art Style

azoomer style / NoobAI, PonyXL, SDXL, SD 1.5 

v4.0

SD 1.5 v4

Files with SHA256: 602164b23e2128f547245823e4becc3eddfc801d0358d21e2070a8b5d26ec7cd