超超超超超好用的flux upscale(使用florence提示词反推)图片放大+细节增加
5.0
0 reviewsDescription
强烈建议使用“4xNomos8kSCHAT-L.pth”这个放大模型,这个模型使用gpu推理放大,放大结果还原原图,就算单用也非常合适。
整个流程使用xl-tile进行第一次放大,为整体图像添加/修改细节,并且进行颜色修正,但是初次放大的图细节增多了但是逻辑性并不强,所以使用flux进行一次整体细节修复,得到了一个非常好的效果
如果需要下载模型:
通过网盘分享的文件:openart模型
链接: https://pan.baidu.com/s/1yamUVEq3gShWdD-NMROGBw 提取码: gjiv
--来自百度网盘超级会员v2的分享
It is strongly recommended to use the "4xNomos8kSCHAT-L.Pth" zoom in model, which uses GPU inference to zoom in and restore the original image. Even if used alone, it is very suitable.
The entire process used XL file for the first zoom in, adding/modifying details to the overall image and performing color correction. However, the details of the first zoomed in image increased but the logic was not strong. Therefore, Flux was used for an overall detail repair, which achieved a very good effect
If you need to download the model:
File shared through online storage: openart model
Link: https://pan.baidu.com/s/1yamUVEq3gShWdD-NMROGBw Extraction code: gjiv
--Share from Baidu Netdisk Super Member v2
Discussion
(No comments yet)
Loading...
Reviews
No reviews yet
Versions (1)
- latest (3 months ago)
Node Details
Primitive Nodes (12)
Anything Everywhere3
Florence2ModelLoader
Florence2Run
FluxGuidance
Image Comparer (rgthree)
ModelSamplingFlux
UnetLoaderGGUF
easy getNode
easy setNode
Custom Nodes (32)
- DeepTranslatorTextNode
- CR Upscale Image
- CR Text Concatenate
ComfyUI
- CheckpointLoaderSimple
- ControlNetApply
- CLIPTextEncode
- VAELoader
- LoadImage
- VAEEncode
- VAEDecode
- PreviewImage
- ControlNetLoader
- ImageScaleBy
- DualCLIPLoader
- UNETLoader
- SamplerCustomAdvanced
- KSamplerSelect
- BasicScheduler
- BasicGuider
- RandomNoise
- SaveImage
- EmptyLatentImage
- KSamplerAdvanced
- easy imageColorMatch
- TextInput_
Model Details
Checkpoints (1)
forrealxl_v05.safetensors
LoRAs (0)