![HYPER SUPIR Upscaling Pack [UPDATED]版本v2.0-hyper (ID: 500770)](https://image.1111down.com/xG1nkqKTMzGDvpLrqFT7WA/9952ea88-6c92-47a4-aa8f-ac87c0c5c44a/width=450/12207169.jpeg)
HYPER SUPIR update
v2.0 Hyper SUPIR video:
v1.0 Lightning SUPIR video:
What is SUPIR? https://github.com/Fanghua-Yu/SUPIR
This SDXL upscaler takes a while, but might offer some fine details to your Upscaling workflow.
be sure your ComfyUI and related custom nodes are up to date ;)
What's in the Pack?
V2.0-hyper
Workflows added for img2img with and without control net.
All workflows from v1.0 updated to use Hyper SDXL 8 step Lora.
v1 pack is included in v2.
V1.0-lightning
txt2img
- Basic-SUPIR
txt2img into SUPIR upscaling, without the triple latent generation from my Trio workflows.
- Trio-SUPIR
Triple latent with SUPIR upscale
- cosTrio-SUPIR
Triple Latents using cosXL, then non-cosXL model for the upscale
- TrioFusionHD-Plus-SUPIR
Triple Latents into IpadapterV2, creates an iterative upscale at the same time as the SUPIR output
img2img
- img2img-SUPIR
takes an image with a simple prompt to upscale using SUPIR, vastly improved in the V2.0 pack.
What extra models do i need?
SUPIR pruned checkpoint:
https://huggingface.co/Kijai/SUPIR_pruned/resolve/main/SUPIR-v0Q_fp16.safetensors?download=true
8 step Hyper Lora (v2.0)
8 step Lightning Lora (v1.0)
Alternate SUPIR profiles are disabled in the workflow, providing a starting point for using SUPIR with or without Lightning.
描述:
Hyper SDXL, used 8 step lora/checkpoint merges
Now with img2img &/OR ControlNet
训练词语:
名称: hyperSUPIRUpscalingPack_v20Hyper.zip
大小 (KB): 76
类型: Archive
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success