<h3 id="settings-qwen-image-:">Settings - Qwen Image :</h3>CFG: 6-8Scheduler: KarrasStrength: 0.5-0.8Sampler: DPM 2Steps : 20-24<h3 id="settings-z-image-turbo-:">Settings - Z-Image Turbo :</h3>CFG: 7Scheduler: BasicStrength: 0.6-0.9Sampler: EulerSteps : 24-28<h3 id="settings-hi-dream-:">Settings - Hi-Dream :</h3>CFG: 5Scheduler: (included with sampler)Strength: 0.7-1.0Sampler: DPM 2 ++ KarrasSteps : 30<h3 id="settings-flux-1-dev-:">Settings - Flux 1 - Dev :</h3>CFG: 4Scheduler: (included with sampler)Strength: 0.4-0.6Sampler: DPM 2Steps : 30<h3 id="settings-sdxl-:">Settings - SDXL :</h3>CFG: 7Scheduler: KarrasStrength: 0.7-1.0Sampler: DPM 2Steps : 24-28Bring the nostalgic charm of late-night instant photography to your generations with Midnight Hour. Trained on various base models, this LoRA captures the aesthetic of authentic hipster-inspired Polaroid selfies—soft flash, vintage color science, cozy low-light moods, natural poses, and that imperfect analog feel that makes instant photography timeless.Whether you're creating lifestyle portraits, fashion editorials, or social media-inspired photography, Midnight Hour adds a subtle cinematic realism with beautiful vintage vibes.<h2 id="if-you-enjoy-this-model-check-out-my-other-free-and-premium-ai-resources-(workflowsnodeslorasdoras):">If you enjoy this model, check out my other free and premium AI resources (workflows/nodes/loras/doras):</h2><h3 id="ko-fi:-https:ko-fi.comtensorvizion-patreon:-https:patreon.comtensorvizion">🎨 Ko-fi: <a target="_blank" rel="ugc" href="https://ko-fi.com/liquidvizion">https://ko-fi.com/tensorvizion</a> ⭐ Patreon: <a target="_blank" rel="ugc" href="https://patreon.com/liquidvizion">https://patreon.com/tensorvizion</a></h3>

MidnightHour

Trained on Qwen 2512

MidnightHour-Qwen-2_c1-st4000.safetensors

Qwen 2512

Qwen Image 1

Flux 1 Dev

HiDream

Z-Image-Base


# 

## Model Introduction

This model is trained using the [Civision/model training](https://modelscope.ai/civision/modelTraining) environment and computing resources provided by ModelScope.

* Model Type: LoRA
* Base Model: [Qwen/Qwen-Image-2512](https://modelscope.ai/models/Qwen/Qwen-Image-2512)
* Training Code: [DiffSynth-Studio](https://github.com/modelscope/DiffSynth-Studio)
* Data amount：26
* Total training steps：4000
* License: Apache-2.0

## Inference Code

Install [DiffSynth-Studio](https://github.com/modelscope/DiffSynth-Studio):

```bash
pip install diffsynth
```

Inference:

```python
from diffsynth.pipelines.qwen_image import QwenImagePipeline, ModelConfig
import torch

pipe = QwenImagePipeline.from_pretrained(
    torch_dtype=torch.bfloat16,
    device="cuda",
    model_configs=[
        ModelConfig(model_id="Qwen/Qwen-Image-2512", origin_file_pattern="transformer/diffusion_pytorch_model*.safetensors"),
        ModelConfig(model_id="Qwen/Qwen-Image", origin_file_pattern="text_encoder/model*.safetensors"),
        ModelConfig(model_id="Qwen/Qwen-Image", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
    ],
    tokenizer_config=ModelConfig(model_id="Qwen/Qwen-Image", origin_file_pattern="tokenizer/"),
)
pipe.load_lora(pipe.dit, ModelConfig(model_id="tensorvizion/MidnightHour-Qwen-2", origin_file_pattern="MidnightHour-Qwen-2_c1-st4000.safetensors"))
prompt = "a cat"
image = pipe(prompt)
image.save("image.jpg")
```


Files with SHA256: aa8a792ef7ba723b9beb811e880c4f85235844e3e0a7119fb96cd32c8e05c295