Oldify

Model Type: LoRA
Trained On: Flux Kontext (via AI Toolkit)
Best Results With: Flux Dev (T2I / I2I modes)
Trigger Word: change the age
Purpose: Transform youthful faces into realistic aged versions, while preserving identity and lighting

---

🧠 What is Oldify?

Oldify is a generative LoRA designed to simulate natural human aging. It introduces realistic age markers—wrinkles, sagging skin, volume loss, and hair graying—while preserving the original subject’s identity, lighting, and composition.

The model was trained using 209 pairs of images generated with StyleGAN3, using age boundary manipulation to transform young faces into older counterparts with consistent facial structure and lighting.

---

To unlock samplers like ipndm and schedulers like kl_optimal, install the custom samplers node: 🔗 https://github.com/ClownsharkBatwing/RES4LYF

Example Prompt:

change the age to around 60 years old with natural signs of aging such as white or gray hair and gentle facial wrinkles, maintaining original skin tone, facial structure, expression, and lighting, realistic aged skin texture with pores and fine lines, high detail and photorealistic quality

⚠️ Important: This LoRA is designed to make people look older, not younger. It adds realistic signs of aging and is not intended for rejuvenation.

---

🎯 What Oldify Changes

Oldify introduces visible but subtle aging effects, including:

• Forehead wrinkles, crow’s feet, and under-eye lines

• Nasolabial folds and slight cheek sag

• Desaturation or graying of hair

• Changes in skin tone and facial volume

• Natural skin texture aging without filter artifacts

---

🧪 Training Info

• Trainer: AI Toolkit

• Steps: 5,000

• Control Dataset: 209 youthful faces

• Target Dataset: 209 aged versions (StyleGAN3 with age boundary edits)

• LoRA Rank: 48

• Resolution: 1024px

• Optimizer: Automagic

• Style: Realistic photographic portraits

• Diversity: Includes various genders and ethnicities

---

⚙️ Usage Guide

🔸 Flux Dev (Recommended)

• Modes: T2I or I2I

• Prompt: Add change the age

• LoRA Weight: 1.2

• Sampler: ipndm

• Scheduler: kl_optimal

• Steps: 20

• Denoise (I2I): 0.45 – 0.65

🔸 Flux Kontext

• Prompt: change the age

• LoRA Weight: 1.2

• Denoise: 1.0

• Sampler: ipndm

• Scheduler: kl_optimal

• Steps: 20

---

🧪 Experimental: Negative LoRA Use

Oldify can optionally be used as a negative LoRA (weight -1.0) to reduce aging or retain a youthful appearance.
In my tests, the best outcome was achieved using:

• Sampler: res_2s

• Scheduler: exponential

However, the result produced unusual artifacts, such as inconsistent skin tone or subtle distortions.
This use case is experimental and not recommended for polished output without visual review.

---

📥 How to Use

Place the .safetensors file in your ComfyUI LoRA folder.
Add change the age to your prompt.
Use the recommended settings for best results.
No extra setup needed — just plug and play.

---

📢 Feedback Welcome!
Feel free to share results, bug reports, or suggestions — it helps improve future LoRAs and ensures better compatibility across pipelines.