# Loras and trigger words

## QI-2512-Emotherapy.safetensors
`https://www.modelscope.ai/models/smansev/emotherapy/summary?version=20260316163648`



#### 1. Minimalist Usage (Baseline)
To get the most consistent and aesthetically pleasing therapy card, use the minimal trigger:

```
EMOTHERAPY happy
EMOTHERAPY brave
EMOTHERAPY frustrated
```

#### 2. Advanced Customization
Break the defaults by injecting specific descriptors for shapes, environments, and lighting.

<img src="https://cdn-uploads.huggingface.co/production/uploads/63e1968e1e5a4f34b7a4b892/sAS90JTIUTRxm9P7b72ZD.png" width="400" />

```
Target Result	Prompt
------------------------
Custom Shape	EMOTHERAPY excited, circular card
Custom Context	EMOTHERAPY lonely, crowded place
Custom Lighting	EMOTHERAPY proud, bright yellow solid background, sunlit rays
Full Override	EMOTHERAPY angry, circular card, warm blue gradient background, school classroom
```

Recommended Inference Settings
For high-fidelity stylization, we recommend these parameters:

Parameter	Recommended Value
Base Model	Qwen-Image
Resolution	1024 x 1024 (Strict 1:1)
LoRA Strength	0.85 - 1.0
Steps	20 - 25
CFG Scale	3.5 - 4.5
Sampler	Euler / FlowMatch

## QIE-2511-SpriteForge-v3.safetensors
`https://www.modelscope.ai/models/tantk7/spriteforgev3/summary?version=20260318050015`

`replace the character in the sprite sheet with the character from the reference image`

<img src="https://cdn-uploads.huggingface.co/production/uploads/63e1968e1e5a4f34b7a4b892/AbiXa1UHthR60AEVrZ7jB.jpeg" width="400" />

Sampler Method: Euler
Guidance Scale: 4
NumInference Steps:30