Model Details
Developed by: AIL-tech
Model type: Diffusion-based text-to-image generative model
Model Description: AIL-Rendering-sd1.5-v10 is the first version of this model, fine-tuned for 600k steps. It can generate high-quality renderings more accurately from tag prompts.
License: CreativeML Open RAIL++-M License
Fine-tuned from: SD 1.5
Model Version
V1.2
The model V1.2 underwent secondary fine-tuning using a training set of 1024*1024, enhancing its ability to generate large images. Recommended output sizes include 1024 x 1024, 1088 x 960, 1152 x 896, 1216 x 832, 1280 x 768, 1344 x 768, 1408 x 704, 1472 x 704, 1536 x 640
Special Tags
Quality tags include: masterpiece, best quality, high quality, good quality, normal quality, low quality, worst quality.
Recommended settings
To guide the model towards generating high-aesthetic images, use negative prompts like:
lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry
For higher quality outcomes, prepend prompts with:
((masterpiece, best quality)),
it’s recommended to use a lower classifier-free guidance (CFG Scale) of around 6-8, sampling steps below 30, clip skip set to 2, and to use Euler Ancestral (Euler a) as a sampler.
Multi Aspect Resolution
This model supports generating images at the following dimensions:
1024 x 1024, 1088 x 960, 1152 x 896, 1216 x 832, 1280 x 768, 1344 x 768, 1408 x 704, 1472 x 704, 1536 x 640
The Future Plan:
This model is the first attempt at training an architecture-specific model, and a SDXL version will be released subsequently."
模型细节
开发者: AIL-tech
模型类型：基于扩散的文生图模型
模型描述：AIL-Rendering-sd1.5-v10是本模型的初代版本，经过了600k步的微调训练。可以更精准的从文本提示生成高质量的效果图。
许可证：CreativeML Open RAIL++-M License
微调自：SD 1.5
模型版本
V1.2
该版本使用1024*1024的训练集进行了二次微调，改善了模型生成大图的能力。推荐生成尺寸为1024 x 1024, 1088 x 960, 1152 x 896, 1216 x 832, 1280 x 768, 1344 x 768, 1408 x 704, 1472 x 704, 1536 x 640
特殊标签
质量标签：masterpiece, best quality, high quality, good quality, normal quality, low quality, worst quality.
推荐设置
为了引导模型生成高审美的图像，可以使用负面提示词：
lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry
为了获得更高质量的结果，可以使用提示前置：
((masterpiece, best quality)), highres,
建议使用约为6-8的（CFG Scale），采样步骤低于30，clip skip设置为2，并使用（Euler a）作为采样器。可以尝试使用DPM adaptive，速度较慢，但可能会获取更细腻的效果。
分辨率
该模型支持生成以下尺寸的图像：
1024 x 1024, 1088 x 960, 1152 x 896, 1216 x 832, 1280 x 768, 1344 x 768, 1408 x 704, 1472 x 704, 1536 x 640
未来计划
本模型为建筑类模型训练的第一步尝试，后续会推出SDXL版本。
Description
The model underwent secondary fine-tuning using a training set of 10241024, enhancing its ability to generate large images. Recommended output sizes include 10241024, 1344*768, 1472*704, and 1536*640.
使用1024*1024的训练集进行了二次微调，改善了模型生成大图的能力。推荐生成尺寸为1024*1024,1344*768，1472*704,1536*640.