Dan Dan is daydreaming: ancient landscapes - Chinese architecture - aerial mirror art - Tencent's Huanyuan Grand Model -

Dandan daydreaming: Ancient Chinese architecture|Strengthening Chinese elements|Enhancing visual style|Designing scenes with empty shots|Tencent Hybrid DiT_V1.2 model

Model introduction: Particularly suitable for creating original materials for Chinese ancient style images, novel promotion, artistic illustrations, and graphic videos.

Creative overview: Dandan has recently been obsessed with traditional Chinese painting style, Chinese elements, and ink rendering.

The more we look at things in China, the more charming they seem.

Model usage instructions: The checkpoint membrane is based on Tencent Hybrid iT_v1.2 large model training.

When testing this Lora small model, other second-dimensional large models, three-dimensional large models, and realistic human large models are also tested for image generation, all producing excellent ink effects.

Prompts usage precautions: Trigger words are dandan, Chinese,

However, the effectiveness of scenes with male characters as the main focus has not been thoroughly tested yet.

Prompts  eg:

raining,natural scenery,green tree,spring,hills,Ancient Chinese palaces,cinematic_angle,depth of field,fisheye,dynamic_angle,lens_flare,landscape,dandan,real,

HDR,UHD,8K,best quality,masterpiece,Highly detailed,Studio lighting,ultra-fine painting,sharp focus,physically-based rendering,

extreme detail description,Professional,masterpiece,best quality,delicate,beautiful,

Other related parameter settings recommendations:

1. Sampling method: Recommending DPM++2M Karras, Eula, Eula A, SDE, with iteration steps of 20-30. Other sampling methods such as LCM, with iteration steps ≤15, can also be tried.

2. Enable high-resolution repair: For realistic human large models, it is recommended to use 8x-NMKD-Superscale, 4x-UltraSharp enlargement algorithms; for second-dimensional, 3D large models, R-ESRGAN_4x+ Anime6B, R-ESRGAN_4x+, and similar enlargement algorithms can be used.

3. Test image resolution sizes: 1:1, 9:16, 16:9, 3:4, 4:3, 512*512, 1024*1024; or 512*768; or 768*1024, 600*800, etc.

4. It is suggested to enable: If drawing characters, enable After Detailer facelift plugin, face_yolov8m, face_yolov8n, face_yolov8n_v2, or multiple units, trigger hand parts with hand_yolov8n, hand_yolov8s, etc.

Welcome all creators to use, provide feedback, and communicate in a friendly manner.