# InkLux v1Watercolor and ink washes with an editorial-illustration sensibility, over clean anime linework. Soft pigment bleed and visible paper texture, but with the deliberate composition and negative space of a magazine spread rather than a straight character portrait.Works on characters, still life and landscapes. It leans painterly, so it does best when you give it room — full-body shots, wide framing, simple backgrounds.## How to useWeight: <code>1.0</code>. The internal balance between the three components is already baked in, so the slider scales all of them together. Lower it to <code>0.6–0.8</code> if the watercolor texture is fighting your character LoRA.Trigger: <code>wc_painting, elstyle</code>If you also grab the companion embedding, <code>embedding:inklux</code> does exactly the same thing in one word — it is a literal copy of those six token embeddings, not an approximation, so the two are interchangeable and produce identical conditioning.There is a third style component with no trigger of its own; it is always active while the LoRA is loaded. That is intentional and it is what carries the linework.Recommended checkpoints: Illustrious-lineage models. Built against Illustrious XL v2.0. It will run on plain SDXL, but two of the three components were trained on Illustrious derivatives, so expect it to be weaker there.Suggested starting prompt:```masterpiece, best quality, wc_painting, elstyle, 1girl, solo, long hair,white shirt, sitting, window, sunlight, simple background```## What's insideA merge of three SDXL LoRAs:- AquaInk – Watercolor Style @ 1.0 — the watercolor body and pigment behavior- editorialluxstyle v3 @ 1.0 — editorial composition, negative space, sketch quality- 96YOTTEA-WAI @ 0.30 — anime linework and character renderingThe merge is a rank-space concatenation, not an SVD approximation: the resulting delta is mathematically identical to stacking all three LoRAs at those weights in your workflow. Nothing was resampled or compressed, which is why the file is larger than a normal rank-32 LoRA (rank 80 where all three overlap).One consequence worth knowing: you can no longer rebalance the three components individually. If the anime linework is too strong for what you want, that ratio is fixed at merge time.## CreditsAll credit for the actual training goes to the authors of the three source LoRAs listed above. This upload is only a merge and contributes no new training.

Files with SHA256: b4e5c35a3dc9231aa1ad9b935b8466d811ffe6d8813e25186c88166b0f4cff2c