# 📷 Analog Style — SDXL LoRA | Zavychroma XLA style LoRA trained on 465 real analog film photographs scraped fromthe film photography community, covering a wide range of cameras, film stocks,subjects, and lighting conditions. It pushes Zavychroma XL toward authenticanalog aesthetics: organic grain, warm or faded tones, soft contrast, scannerartifacts, and the beautiful imperfections of real film.---## ℹ️ Model Info| Type | LoRA (UNet-only, attention layers only) || Base model | Zavychroma XL v1.0 (SDXL) || Rank / Alpha | 16 / 16 || Training precision | FP16 || Final loss | ~0.008 |&gt; ⚠️ Trained on Zavychroma XL. Use with Zavychroma or close SDXL merges&gt; for best results. Other SDXL checkpoints may work but results will vary.---## 🎯 Trigger Wordanalogstyle Always include <code>analogstyle</code> in your positive prompt.---## 📷 What the model learned fromThe dataset included real scans from:Cameras: Leica M2/M6/MP, Mamiya RZ67/RB67/7/7ii, Hasselblad 500C/CM,Rolleiflex 2.8F/3.5F, Nikon F2/F100/FM2/F6, Canon A1/AE1/EOS, Pentax 67/67ii,Contax G1/G2, Minolta X700, Olympus XA/OM4ti, and more.Film stocks: Kodak Portra 160/400/800, Gold 200, Ultramax 400, Ektar 100,CineStill 800T/50D, Fuji Superia 400/800, Velvia 100, Provia, Natura 1600,Ilford HP5, Ilford Delta, Kodak T-Max, Aerochrome, Lomo Redscale, and more.Subjects: portraits, street photography, travel, nature, landscapes, weddings,architecture — shot in Japan, Paris, Seoul, Norway, Morocco, San Francisco,Mongolia, Italy, and more.---## ✅ Recommended Settings- Sampler: DPM++ 2M Karras or Euler a- Steps: 25–35- CFG: 6–7.5- Resolution: 1024×1024 or 832×1216 / 1216×832- LoRA weight: 0.7–1.0---## 💡 Prompting Tips- The trigger <code>analogstyle</code> alone already shifts the aesthetic — dial the LoRA weight up/down to control intensity- Name a film stock to push toward that specific look: <code>Kodak Portra 400</code>, <code>CineStill 800T</code>, <code>Fuji Velvia 100</code>, <code>Ilford HP5</code>- Describe light explicitly: <code>golden hour</code>, <code>window light</code>, <code>overcast soft light</code>, <code>neon street light</code>- For portraits add: `shallow depth of field, slight lens aberration, film grain, scanned negative`- For black and white: <code>black and white, Ilford HP5, high grain, deep blacks</code>---## 🧪 Test PromptsPrompt A — Portrait, warm light:analogstyle, portrait of a young woman near a window, Kodak Portra 400, golden hour, shallow depth of field, film grain, slight vignette, 35mm lens, candidNegative: digital, HDR, oversaturated, airbrushed, CGIPrompt B — Street, tungsten night:analogstyle, street scene in a European city at night, CineStill 800T, tungsten glow, grain, halation on highlights, documentary photography, 28mm lensNegative: digital sharp, clean, HDR, oversaturated---## 📌 Technical Notes- UNet-only — text encoder untouched, so prompts fully control subject and composition- Attention-only (Q/K/V/Out) — clean style transfer without distorting spatial layout- Full depth coverage of <code>down_blocks_2</code>, <code>mid_block</code>, and <code>up_blocks_0</code> = strong style influence even at moderate weights

Files with SHA256: d5403495153c67984138a224c3b7592bd577dc64519d088473ab8311e24a2491