A niche, high‑precision concept for creators who need real partner dynamics.Partner Flow — Tango is a LoRA designed to teach the model the actual grammar of tango: balance, connection, posture, and expressive movement between two dancers. This is not a generic “dance LoRA”. It’s a specialized concept, intentionally niche, built for artists who want cinematic, elegant, believable partner poses.It requires a bit of prompt discipline — but when used correctly, it delivers film‑quality scenes.<h2 id="what-this-lora-does">🎭 What this LoRA does</h2>Partner Flow — Tango focuses on:<ul><li>Authentic tango poses (close‑position, open‑position, cross‑steps, ochos, ganchos, casqué, dramatic dips)</li><li>Partner interaction (support, weight distribution, hand placement, body alignment)</li><li>Clean leg lines (no fusions, no distortions, stable footwork)</li><li>Cinematic atmosphere (night streets, wet pavement, fog, depth, rim light)</li><li>Lighting coherence (warm streetlights, cool moonlight, mixed lighting)</li><li>Strong generalization (different outfits, colors, lighting setups, environments)</li></ul>The LoRA doesn’t “copy” the dataset — it understands the concept and adapts it to new contexts.<h2 id="dataset-structure">📂 Dataset Structure</h2>The dataset was built in four curated folders, each with a specific purpose:<h3 id="1)-static-poses">1) Static Poses</h3><ul><li>Close‑position tango</li><li>Balanced, elegant posture</li><li>Teaches stability and body alignment</li></ul><h3 id="2)-dynamic-poses">2) Dynamic Poses</h3><ul><li>Turns, ochos, ganchos, lateral steps</li><li>Teaches movement, flow, and leg precision</li></ul><h3 id="3)-night-environments">3) Night Environments</h3><ul><li>Wet streets</li><li>Streetlamps</li><li>Atmospheric fog</li><li>Teaches cinematic composition and depth</li></ul><h3 id="4)-lighting">4) Lighting</h3><ul><li>Warm golden light</li><li>Cool moonlight</li><li>Mixed lighting</li><li>Teaches mood, contrast, and reflection behavior</li></ul>The dataset is clean, coherent, and stylistically unified, which is why the LoRA generalizes so well.<h2 id="recommended-lora-weight">⚙️ Recommended LoRA Weight</h2>0.8 – 1.0<ul><li>0.8 → more creative freedom, softer influence</li><li>1.0 → full tango grammar, sharper posture and lines</li></ul>Both ranges are stable and produce consistent results.<h2 id="trigger-word">🔑 Trigger Word</h2><code>pf_tango</code>Not mandatory, but strongly recommended to activate the full concept.<h2 id="notes">📸 Notes</h2>Partner Flow — Tango is a professional‑grade LoRA aimed at creators who want:<ul><li>expressive dance scenes</li><li>believable partner interaction</li><li>cinematic night atmospheres</li><li>elegant storytelling through movement</li></ul>It’s a niche concept — not “plug &amp; play” — but when used with intention, it produces striking, emotionally rich images.

PartnerFlow

PF Bachata is a high‑precision LoRA designed to capture the essence of Latin dance, with a strong focus on bachata, couple dynamics, pre‑dance tension, and editorial‑grade lighting.Although the concept is natively Hispanic/Colombian, the model demonstrates exceptional generalization across:<ul><li>Ethnicities (Latina, Afro‑Latina, African, mixed, Mediterranean)</li><li>Outfits (modern, traditional, African‑inspired, Caribbean, fashion, invented)</li><li>Environments (Latin America, Caribbean, Spain, Africa, studio, abstract)</li><li>Lighting styles (studio hard light, rim light, neon, golden hour, low‑key, high‑key)</li></ul>PF Bachata does not simply reproduce the dataset — it understands the visual language of dance and adapts it to new contexts with remarkable consistency.🎯 Recommended Strength0.7 – 1.0The LoRA is stable across this entire range. Even at lower weights it maintains identity, pose structure, and lighting behavior.🧩 Core TagsThe model was built using a tag‑first methodology, defining the visual grammar before generating the dataset. This results in extremely clean, predictable behavior.Identity Tags<ul><li><code>pf_bachata</code> (primary trigger)</li><li><code>colombian latin dancers</code></li><li><code>afro-colombian dancers</code></li></ul>Base Poses<ul><li><code>pose_open_step</code></li><li><code>pose_close_embrace</code></li><li><code>pose_cross_body</code></li><li><code>pose_turn</code></li></ul>Dynamic Poses<ul><li><code>pose_sensual_wave</code></li><li><code>pose_leg_hook</code></li><li><code>pose_dip</code></li><li><code>pose_body_isolation</code></li></ul>Pre‑Dance Variants<ul><li><code>pose_open_step_pre</code></li><li><code>pose_close_embrace_pre</code></li><li><code>pose_sensual_wave_pre</code></li><li><code>pose_cross_body_pre</code></li><li><code>pose_body_isolation_pre</code></li></ul>These tags allow the model to maintain dance logic, even in completely new settings.🌍 Ethnicity GeneralizationDespite being trained on a Hispanic/Colombian concept, PF Bachata generalizes far beyond its native domain:<ul><li>Caramel‑skinned Colombian women and men</li><li>Afro‑Colombian dancers</li><li>African dancers with dreadlocks, natural curls, or shaved styles</li><li>Mixed‑ethnicity couples</li><li>Mediterranean and Caribbean looks</li></ul>The model preserves:<ul><li>facial structure</li><li>cultural coherence</li><li>skin tone accuracy</li><li>dance dynamics</li></ul>…even when pushed into non‑Latin environments like African savannas or Spanish courtyards.👗 Outfit GeneralizationPF Bachata handles clothing with impressive flexibility:<ul><li>respects the original dataset’s dancewear</li><li>adapts to new dress cuts</li><li>accepts new color palettes</li><li>supports African‑inspired outfits</li><li>supports modern fashion</li><li>supports invented or editorial clothing</li><li>maintains dance‑appropriate silhouettes</li></ul>The model does not overfit to specific garments — it interprets the outfit while keeping the bachata identity intact.💡 Lighting &amp; Photography BehaviorOne of the strongest aspects of PF Bachata is its editorial‑grade lighting intelligence.It responds beautifully to:<ul><li>hard studio light</li><li>rim light / edge light</li><li>low‑key cinematic</li><li>high‑key fashion</li><li>neon nightclub lighting</li><li>golden hour</li><li>dramatic shadow sculpting</li><li>soft diffused portrait light</li></ul>Lighting changes mood, emotion, and tension in a natural, photographic way.🧠 Model Behavior &amp; StrengthsPF Bachata consistently demonstrates:<ul><li>strong couple interaction logic</li><li>believable pre‑dance tension</li><li>sensual but controlled expressions</li><li>clean hands and body proportions</li><li>stable half‑body and portrait performance</li><li>excellent full‑body dance dynamics</li><li>robust generalization to new ethnicities</li><li>robust generalization to new outfits</li><li>robust generalization to new environments</li><li>consistent editorial quality</li><li>cinematic mood control</li><li>high detail retention even at low weights</li></ul>This LoRA behaves more like a mini‑checkpoint than a typical LoRA.📸 ShowcaseThe model includes an extensive showcase featuring:<ul><li>portraits</li><li>half‑body</li><li>full‑body</li><li>pre‑dance tension</li><li>dynamic dance poses</li><li>mixed ethnicities</li><li>varied lighting</li><li>varied outfits</li><li>varied environments</li></ul>All showcase images include their original prompts, which users can reference for guidance.📝 Usage Notes<ul><li>Always include <code>pf_bachata</code> to activate the LoRA.</li><li>For dance scenes, specify the pose tag.</li><li>For pre‑dance tension, use the <code>_pre</code> variants.</li><li>For portraits, use half‑body or close‑up framing.</li><li>Ethnicity can be changed simply by describing it.</li><li>Outfits can be freely invented.</li><li>Lighting dramatically influences mood — use it intentionally.</li><li>Works extremely well with cinematic, editorial, and fashion‑style prompts.</li></ul>

Files with SHA256: 8c138e752dc91f7acdf8792f42c6245eae0980e62101923a9454a43e1333cdf1