<h1 id="version-3-is-a-full-rebuild-not-a-tweak-of-halo.">Version 3 is a full rebuild, not a tweak of Halo.</h1>The checkpoint was reconstructed from the ground up on Lumen. Halo’s weights are not the base of this release. Style has been updated across the model: cleaner rendering, a more consistent look, and a different finish from the Halo version.What changed<ul><li>The complete rebuild is on Lumen, not a merge on top of Halo.</li><li>The style pass is updated.</li></ul>This is a new version, not a patch of the previous checkpoint. XavierNocturne v2 · HaloIllustrious · furry / anthro Follow-up to v1.1 Lumen. Not a cosmetic bump.<h2 id="what-it-is">What it is</h2>Lumen stays a night checkpoint with a volumetric beam. Halo does not add a new paint job. It adds hearing: stronger frame and camera follow-through without throwing the light away.<h2 id="what-was-merged">What was merged</h2><ul><li>Part Source Role UNet OUT, finish, VAE XavierNocturne v1.1 Lumen light, fur, night, color UNet IN + part of MID WAI-Illustrious-SDXL v14, pose, camera, layout CLIP-L + CLIP-G ESP1 prompt </li></ul>WAI14 was not used as a full replacement checkpoint. No lighting LoRA baked into the core.<h2 id="blocks">Blocks</h2>SDXL reads: CLIP → IN → MID → OUT.<ul><li>IN00–IN08 — who stands where, and from which camera.</li><li>MID — scene glue.</li><li>OUT00–OUT08 — light, material, Lumen’s face.</li><li>CLIP — how words reach IN. XavierNocturne v1.1 Lumen - Update Notes</li></ul>Version Update: Improved Lighting &amp; Refined AnatomyThank you for all the support, downloads, and feedback. This update focuses on two major areas that users requested the most: lighting quality and anatomical consistency.What’s New<ul><li>Significantly improved light response and volumetric behavior</li><li>Better subsurface scattering and rim light handling</li><li>More stable and natural anatomy (especially hands, torso, and facial structure)</li><li>Cleaner rendering of fur and skin transitions</li><li>Improved overall coherence in complex scenes</li></ul>Recommended Positive Prompt Basetext<pre><code>masterpiece, best quality, absurdres, very aesthetic,</code></pre>You can freely expand from this base depending on the subject.Lighting Control GuideLumen responds very well to directional and atmospheric lighting prompts. Here’s how to guide the light:Soft / Natural Lighttext<pre><code>soft lighting, diffused light, gentle volumetric lighting</code></pre>Dramatic / Cinematic Lighttext<pre><code>dramatic lighting, strong rim light, volumetric lighting, cinematic lighting</code></pre>Backlight / Rim Lighttext<pre><code>backlit, strong rim light, light from behind, glowing edges</code></pre>Side Lighttext<pre><code>side lighting, light from the left, light from the right, dramatic side light</code></pre>Top / Overhead Lighttext<pre><code>overhead lighting, light from above, god rays</code></pre>Warm vs Cool<ul><li>Warm: warm lighting, golden hour, soft orange light</li><li>Cool: cool lighting, blue hour, moonlight</li></ul>Combining volumetric lighting + rim light usually gives the most cinematic results.Recommended Negative Prompttext<pre><code>low quality, worst quality, bad anatomy, bad hands, missing fingers, extra digits, mutated hands, poorly drawn face, blurry, jpeg artifacts, text, watermark, logo, censored, extra limbs, deformed, ugly</code></pre>Recommended Settings<ul><li>Sampler: DPM++ 2M SDE or Euler a</li><li>Scheduler: Karras , normal , beta</li><li>Steps: 28–32</li><li>CFG Scale: 2.8 – 3.5</li><li>Clip Skip: 2</li><li>Resolution: 832×1216 / 1024×1024 / 1216×832 (or higher with upscale)</li></ul>Recommended Settings for CivitAi generator. Sampler: Euler a / DPM++ 2M SDE. Steps: 25–35. CFG: 3.5–5.5. Clip Skip: 2Hires.fix Denoise: 0.3– 0.5. Hires.fix: 1.3–1.5 + R-ESRGAN 4x+ Anime6B, RemacriFree_2U IntegrationLumen works very well with Free_2U (FreeU v2). It improves structure, sharpness, and overall coherence without needing extra steps.Recommended Starting Settings (Balanced)Parameter Value Description b1 1.2 Backbone scale (Stage 1) – controls overall structure and stability b2 1.35 Backbone scale (Stage 2) – affects finer details and contrast s1 0.90 Skip scale (Stage 1) – reduces low-frequency noise in early blocks s2 0.20 Skip scale (Stage 2) – cleans up high-frequency artifactsAlternative PresetsMild (safer, less risk of oversharpening):<ul><li>b1: 1.1</li><li>b2: 1.25</li><li>s1: 0.95</li><li>s2: 0.30</li></ul>Strong (more detail &amp; punch):<ul><li>b1: 1.3</li><li>b2: 1.45</li><li>s1: 0.85</li><li>s2: 0.15</li></ul>Tips<ul><li>Start with the Balanced preset.</li><li>If you notice halos, crunchy edges, or over-sharpened fur/skin — lower b1 and b2 first.</li><li>s2 has a strong effect on fine detail cleanliness. Lower values = cleaner but can sometimes lose softness.</li><li>Free_2U pairs especially well with the improved lighting and anatomy of this version.</li></ul>Post-Processing Tips for a Cinematic LookFor a more film-like and polished result, try this combination after generation:<ol><li>Image Sharpen - light to medium strength (avoid over-sharpening)</li><li>Contrast Fix - subtle increase for better depth</li><li>Image Film Grain - low to medium amount for a natural cinematic texture</li></ol>This trio helps the image feel less “digital” and more like a still from a movie.Thank you again for downloading, testing, and supporting XavierNocturne. Your feedback directly shapes the next updates.If you create something cool with this version — feel free to share it. Enjoy generating.XavierNocturne v1 Indigo/XavierInHell 65/35Merge of Indigo Furry Mix (65%) and Xavier (35%). This model combines the strong anatomy, body proportions, and lighting of Indigo Furry Mix with the distinctive style and aesthetic of Xavier. It delivers solid structure while preserving a unique visual character, making it well-suited for detailed furry and anthro generations.<h3 id="compatibility">Compatibility</h3><ul><li>Fully works in ComfyUI</li><li>Fully works in Automatic1111 / Forge / reForge</li></ul><h3 id="recommended-settings">Recommended Settings</h3>Sampler:<ul><li>euler / euler_a (best)</li><li>dpmpp_2m</li><li>dpmpp_sde</li></ul>Scheduler:<ul><li>normal (best)</li><li>karras</li><li>beta</li></ul>CFG Scale: 3.5 – 6 Steps: 25 – 35 Recommended prompt: Positive: masterpiece, best quality, very aesthetic, detailed eyes, detailed face, intricate details, beautiful lighting, dramatic lighting, volumetric lighting, sharp focus, highly detailed, Negative: lowres, worst quality, bad quality, bad anatomy, bad hands, extra fingers, missing fingers, blurry, low quality, text, watermark, ugly, poorly drawn face, poorly drawn hands,<h3 id="strengths">Strengths</h3><ul><li>Good anatomy and body structure</li><li>Quality lighting and depth</li><li>Distinctive style from Xavier</li><li>Handles detailed fur, clothing, and complex scenes well</li></ul><h3 id="notes">Notes</h3><ul><li>Best results at moderate CFG</li><li>Compatible with most SDXL LoRAs</li><li>Optional FreeU_V2 or ADetailer can further improve details</li><li>Works reliably in both ComfyUI and A1111-based interfaces</li></ul><h3 id="if-you-like-it-post-the-image-to-this-model-gallery">If you like it — post the image to this model gallery <img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/171b3252-1040-410b-b9ff-e640a1dfaaa3/original=true/171b3252-1040-410b-b9ff-e640a1dfaaa3.jpeg" /></h3>

XavierNocturne

xaviernocturne_v3.safetensors

v2 | Halo

v1.1 | Lumen

v1.0

creator: Zapask10IndigoXavier 65/35Merge of Indigo Furry Mix (65%) and Xavier (35%). This model combines the strong anatomy, body proportions, and lighting of Indigo Furry Mix with the distinctive style and aesthetic of Xavier. It delivers solid structure while preserving a unique visual character, making it well-suited for detailed furry and anthro generations.Compatibility<ul><li>Fully works in ComfyUI</li><li>Fully works in Automatic1111 / Forge / reForge</li></ul>Recommended SettingsSampler:<ul><li>euler / euler_a (best)</li><li>dpmpp_2m</li><li>dpmpp_sde</li></ul>Scheduler:<ul><li>normal</li><li>karras</li><li>beta (best)</li></ul>CFG Scale: 3.5 – 6 Steps: 25 – 35 Recommended prompt: Positive: masterpiece, best quality, very aesthetic, detailed eyes, detailed face, intricate details, beautiful lighting, dramatic lighting, volumetric lighting, sharp focus, highly detailed, Negative: lowres, worst quality, bad quality, bad anatomy, bad hands, extra fingers, missing fingers, blurry, low quality, text, watermark, ugly, poorly drawn face, poorly drawn hands,Strengths<ul><li>Good anatomy and body structure</li><li>Quality lighting and depth</li><li>Distinctive style from Xavier</li><li>Handles detailed fur, clothing, and complex scenes well</li></ul>Notes<ul><li>Best results at moderate CFG</li><li>Compatible with most SDXL LoRAs</li><li>Optional FreeU_V2 or ADetailer can further improve details</li><li>Works reliably in both ComfyUI and A1111-based interfaces</li></ul>

Files with SHA256: 40fbb6e187a9a96f895fc60cdc26a337ea17f80dc74b7a1e0142534a2d67a68b