Forge Neo Portable 2.30 (CUDA 13.0, Torch 2.13, Python 3.13.14, SageAttention, Gradio 4.40) + Minimax H3 Supporting

If you want download my build from Hugging Face - https://huggingface.co/datasets/StefanFalkok/Forge_Neo_Portable_torch_2.13.0_cu130_cp313_sageattention_triton

manual and credits to https://github.com/Haoming02/sd-webui-forge-classic/tree/neo

Supports Anima (including 2.9B and 3.8B), Flux 2 Klein, Qwen Image 2.1, Krea 2, Minimax H3

works fine and fast like in ComfyUI

I have installed all necessary extenstions with comfortable job:

My build has Danbooru and e621 autotags with October 2026 data

The next update will be an update when Haoming02 added his native support Minimax H3

Added adetailers, controlnet lllite and upscale ESRGan models

You can generate videos via my Forge Neo Build

Maybe you need install https://huggingface.co/datasets/StefanFalkok/Forge_Neo_Portable_torch_2.13.0_cu130_cp313_sageattention_triton/blob/main/VC_redist.x64%20(2).exe for SageAttention and Kitchen Attention working

_____________________________
Start launch via webui-user.bat

if you need change argument - right click on webui-user.bat, edit in Notepad and you can add arguments in set COMMANDLINE_ARGS=

Choose the preset for certain models

Choose diffusion model, text encoder and vae you need

to enable Edit working, go to setting/Stable Diffusion and turn on it

For Krea2 and Anima you need separate Lora Edits

for Anima - https://civitai.red/models/2650553/anima-edit

for Krea 2 - https://civitai.red/models/2761113/krea-2-identity-edit

Krea 2 Character sheet - https://civitai.red/models/2764727/character-sheets?modelVersionId=3203877

To avoid OOM, click Never OOM intergrated and Enable for Unet