README.md

---
tags:
- text-to-image
- lora
- diffusers
- template:diffusion-lora
widget:
- output:
    url: images/TryAnything_00204_.png
  text: '-'
- output:
    url: images/TryAnything_00205_.png
  text: '-'
- output:
    url: images/TryAnything_00144_.png
  text: '-'
- output:
    url: images/TryAnything_00145_.png
  text: '-'
- output:
    url: images/TryAnything_00158_.png
  text: '-'
- output:
    url: images/TryAnything_00159_.png
  text: '-'
- output:
    url: images/TryAnything_00154_.png
  text: '-'
- output:
    url: images/TryAnything_00155_.png
  text: '-'
- output:
    url: images/TryAnything_00148_.png
  text: '-'
- output:
    url: images/TryAnything_00149_.png
  text: '-'
- output:
    url: images/TryAnything_00128_.png
  text: '-'
- output:
    url: images/TryAnything_00129_.png
  text: '-'
- output:
    url: images/TryAnything_00124_.png
  text: '-'
- output:
    url: images/TryAnything_00125_.png
  text: '-'
- output:
    url: images/TryAnything_00122_.png
  text: '-'
- output:
    url: images/TryAnything_00123_.png
  text: '-'
base_model: black-forest-labs/FLUX.1-Kontext-dev
instance_prompt: try on this outfit, man try on this outfit, woman try on this outfit
license: apache-2.0
---
# Try Anything

<Gallery />

## Model description 

**Model Type:** LoRA  
**Trained On:** Flux Kontext (via AI Toolkit)  
**Best Results With:** Flux Kontext &#x2F; Flux Dev (untested but likely compatible)  
**Trigger Word:** &#x60;try on this outfit&#x60;  
**Purpose:** Virtual clothing try-on for realistic outfit visualization  

**Workflow:** [Available on Patreon](https:&#x2F;&#x2F;www.patreon.com&#x2F;posts&#x2F;try-anything-134558142?utm_medium&#x3D;clipboard_copy&amp;utm_source&#x3D;copyLink&amp;utm_campaign&#x3D;postshare_creator&amp;utm_content&#x3D;join_link)

---

## 🧠 What is TryAnything?

**TryAnything** is a virtual try-on LoRA designed to transfer clothing from product images onto people. It enables realistic visualization of how outfits would look on different individuals while preserving identity.

The model was trained using **240 real-person images** from VTON datasets, focusing primarily on upper body garments with the ability to handle full outfit combinations.

---

## 🎯 What TryAnything Does

TryAnything transfers clothing details including:

- Upper body garments (shirts, blouses, jackets, tops)
- Full outfit combinations (top + bottom)
- Clothing fit and draping
- Fabric textures and patterns
- Natural clothing appearance on the target person

---

## 🧪 Training Info

- **Trainer:** AI Toolkit
- **Steps:** 6,000
- **Dataset:** 240 real-person VTON images
- **LoRA Rank:** 32
- **Resolution:** 1024px
- **Base Model:** Flux Kontext
- **Focus:** Upper body garments with full outfit capability

---

## ⚙️ Usage Guide

### **Recommended Settings:**
- **LoRA Weight:** 1.0 (adjust as needed)
- **Trigger:** &#x60;woman try on this outfit&#x60; or &#x60;man try on this outfit&#x60;
- **Best Practice:** Use clothing images without people wearing them
- **Can work:** With people wearing clothes, but not recommended

### **Example Prompts:**
```
try on this outfit
```
```
try on this outfit, man
```
```
try on this outfit, man, copy t-shirt and shorts
```
```
try on this outfit, blouse with short sleeves
```

---

## ⚠️ Important Notes - Version 1 Limitations

### **Current Limitations:**
- **V1 Status:** This is the first version with room for improvement
- **Head Proportions:** May have issues with head-to-body proportions (most common problem)
- **Training Focus:** Primarily trained on upper body garments
- **Bottom Wear:** Limited training on lower body clothing

### **Best Results:**
- Use standalone clothing images (not worn by people)
- Specify clothing details in prompts when needed
- Include &quot;anatomically correct proportions&quot; in prompts

### **Future V2 Plans:**
- Enhanced training with underwear and foundation garments
- Improved lower body clothing handling
- Better proportion consistency

---

## 📥 How to Use

1. Load the LoRA in your workflow
2. Use trigger: &#x60;woman try on this outfit&#x60; or &#x60;man try on this outfit&#x60;
3. Provide clear clothing reference images (preferably unworn)
4. Include clothing details in prompt if needed
5. Add &quot;anatomically correct proportions&quot; for better results

---

## 🔬 Compatibility

- **Primary:** Flux Kontext
- **Likely Compatible:** Flux Dev (untested)
- **Weight:** Start with 1.0, adjust up&#x2F;down as needed

---

**Feedback Welcome!** This is V1 - share your results and suggestions for improvements in V2!

## Trigger words

You should use `try on this outfit` to trigger the image generation.

You should use `man try on this outfit` to trigger the image generation.

You should use `woman try on this outfit` to trigger the image generation.


## Download model


[Download](/Alissonerdx/TryAnything/tree/main) them in the Files & versions tab.