The new Accelerator Sampler is recommended, 1.6X speed is recommended, and the number of steps should also be added

TeaCache Accelerated Sampler node link:

https://github.com/TTPlanetPig/Comfyui_TTP_Toolset

The model is a huggingface from kijai, don't use fastvideo lora, it will crash the picture, https://huggingface.co/Kijai/HunyuanVideo_comfy/tree/main

The vae Decode(tiled) version has been upgraded for better, please upgrade the comfyui body and replace it.

Added image inference prompt and Upscale

IKohya updated the training code of Hunyuan video, but it is a bit difficult to install and there is no GUI.

Tangtuanzhu TTPlanet temporarily added a GUI to help everyone quickly train and use it, and also made a simple tutorial, so you can get started with it with zero difficulty.

Kohya original code:

https://github.com/kohya-ss/musubi-tuner/tree/main

Tangtuanzhu TTPlanet made his own graphical interface and integration package, just use it without installing and unzipping!

musubi-tuner_with_gui_python_TTP.7z

https://pan.baidu.com/s/1EzlY728byW5pHVd9mDGgXQ?pwd=s2iz

Decompression password: 汤团猪TTPlanet

You can also refer to Tangtuanzhu TTPlanet previous video tutorial to arrange Hunyuan. No special settings are required, the tutorial is universal

Zero foundation, teach you how to use AI to make videos - starting from "zero" -

https://www.bilibili.com/video/BV1RgkhYREix/

My lora can be found on C station:

https://civitai.com/user/ttplanet