Model type: LoRA Base model: FLUX.1 Kontext Dev Version: v1.0 Trigger / trained word: <code>jmouth</code><h2 id="description">Description</h2>JMouth is a focused FLUX Kontext image-editing LoRA for enlarging lips while preserving the rest of the source image.It was trained on paired before/after images where the intended difference was isolated as closely as possible to lip size and volume only. The goal is to produce fuller, larger lips without unnecessarily changing identity, mouth position, expression, face shape, lighting, skin texture, pose, clothing, or background.This is intended as an image-edit LoRA for FLUX Kontext, not a general text-to-image lip style.<h2 id="recommended-prompt">Recommended Prompt</h2><code>jmouth Change the lips to significantly larger, fuller lips.</code>For a more natural / less glossy result:<code>jmouth Change the lips to significantly larger, fuller lips while preserving the original lip color and natural texture. Keep everything else unchanged.</code>For especially natural bare lips:<code>jmouth Change the lips to significantly larger, fuller lips while preserving the original lip color, natural texture, and matte finish. No lipstick or gloss. Keep everything else unchanged.</code><h2 id="recommended-settings">Recommended Settings</h2>LoRA strength: Start at <code>1.0</code>Suggested range:<code>0.7 – 1.1</code><ul><li>Lower strength for a subtler enlargement.</li><li>Around <code>1.0</code> for the intended effect.</li><li>Higher values can be tested for more exaggerated fullness.</li></ul>Kontext guidance: Around <code>3.5 – 4.0</code> is a good starting range.Use your original image as the Kontext control/reference image.<h2 id="what-it-is-designed-to-preserve">What It Is Designed To Preserve</h2><ul><li>Identity</li><li>Face shape</li><li>Mouth position</li><li>Open/closed mouth state</li><li>Expression</li><li>Visible teeth or tongue</li><li>Nose, eyes and eyebrows</li><li>Hair</li><li>Pose and head angle</li><li>Skin tone and texture</li><li>Freckles and pores</li><li>Lighting and exposure</li><li>Clothing and accessories</li><li>Background</li><li>Original photographic character</li></ul>The intended transformation is simply:smaller/thinner lips → larger/fuller lipsEach training example used a paired source and target image. The source contained smaller lips and the target contained enlarged lips, with unrelated image changes minimized as much as possible.<h2 id="notes">Notes</h2>JMouth is deliberately narrow in scope. It is not intended to redesign the face or apply a general beauty treatment.If the output becomes too glossy or makeup-like, explicitly ask Kontext to preserve the original lip color and texture or specify no lipstick, no gloss, natural matte lips.The LoRA can also be combined with more detailed Kontext instructions when you want to control the degree or character of the enlargement.

Files with SHA256: d125ead72d8b4ff70a0735a91d4227c68f6a36c07958f8f52715d483572b3996