Type | |
Stats | 1,109 832 |
Reviews | (91) |
Published | Sep 26, 2023 |
Base Model | |
Training | Steps: 45 Epochs: 1 |
Hash | AutoV2 79F6372EB6 |
Refining a modern urban all-element LORA model, crawling a series of high-quality renderings on the network for classification training to obtain this dan, mainly hoping to help engaged in urban design and architectural design related fields. If you can cooperate with the relevant LoRa and the matching scene, the effect will be doubled with half the effort.
Recommendations for use:
1.It is recommended to adjust the weight between 0.3-0.5, and the effect will explode when you increase it, you can give it a try
2.The recommended sampling method is DPM++ 2S a Karras
3.The view frame is controlled within 900 width and 600 height, which is easier to produce effects
4.The processor selects MLSD to extract the straight line
5. Negative keyword suggestions:16-token-negative-deliberate-neg,realisticvision-negative-embedding
6.checkpoint collocation majicMIXrealistic 麦橘写实_v2威力加强典藏版 to achieve the effect of adjusting the intensity of the style