Detail Slider LoRA | Anima

作者：Shed_The_Skin

👇👀 Read me! 👇👀

This is a bipolar slider for ANIMA. The purpose of this slider is to increase or decrease the amount of rendered detail in a controllable fashion with the LoRA input values. Higher positive LoRA input values result in finer, richer detail: texture on fabric, hair, foliage, wood and metal, engraving and filigree on armour, more going on in the background. Negative values result in the same picture rendered cleaner and simpler, with smoother surfaces and fewer shading steps.

What makes this slider special is that it adds detail to the whole picture without changing what the picture is. Most detail LoRAs quietly change the art medium, zoom out, add clutter, or, worst of all, age every face they touch, because finer rendering on a face reads as wrinkles and creases. This slider was trained against all of those: guards for apparent age, view angle, saturation, clutter, crowding and realism / 3D drift, a style guard so a painting stays a painting, and a face-protected component that keeps detail off faces. The subject, composition, colours and art style stay put while the detail level changes. It works on characters and on pure scenery (no humans) alike.

Supported input values are -2 to 2. ±1 is the cleanest everyday range; +2 is the strong end, where very close-up faces can start to lean a little more realistic. Please be aware that your prompt still sets the base: quality and detail tags you already use (masterpiece, absurdres, intricate details, etc.) are what the slider works off of. The LoRA slider works like an offset in this regard.

No activation prompt. Example prompt (the showcase images use this style; booru tags work best with Anima):

<lora:StS_Detail_Slider_v1.0_ANIMA:1.5>, masterpiece, best quality, amazing quality, very aesthetic, absurdres, highres, score 9, safe, 1girl, solo, mature female, knight, blonde hair, long hair, single braid, braid over shoulder, blue eyes, looking at viewer, serious, closed mouth, armor, plate armor, silver armor, gold trim, pauldrons, gauntlets, blue tabard, red cape, sword, planted sword, hands on hilt, both hands on hilt, hands on own hilt, standing, straight-on, symmetrical pose, cowboy shot, outdoors, castle, battlements, banner, sunset, orange sky, cloud, rim lighting, fantasy

Showcase settings: hosekiLustrousmixAnima v1.0, Euler, Normal, 22 steps, CFG 3, Shift 1, 768x1024 or 1024x1024, Hires 2x with 4x_fatal_Anime_500000_G, 22 hires steps, denoise 0.3. Full parameters are embedded in every example image. For the cleanest slider sweeps, use a non-ancestral sampler (Euler rather than Euler a) and keep the seed fixed.

If you have issues using this LoRA, feel free to send me a message on CivitAI and I will help you out if I can. If you are running into use cases where this LoRA breaks down and/or freaks out, please let me know so I can fix it for future version releases.

LoRA safetensors available for download via this model's page on TensorArt.

Changelog: November 13, 2026 - Initial v1.0 Release