This node enables the use of VAE decoding through HuggingFace's inference endpoints.
For more details, refer to the VAE Decode with Hybrid Inference documentation.
Note: This node does not support SD1.5 or SDXL VAEs, as they typically perform faster on local hardware.
Currently, only Flux and HunyuanVideo are supported.
As HuggingFace expands support for other VAEs, they will be incorporated into this node.
This node does not rely on the diffusers library and has no external dependencies beyond ComfyUI-Core. It is designed to be minimalistic.
For more information on remote VAE decoding, you can refer to HuggingFace's blog post
Using this node is quite simple, just replace the built-in VAE Decode with VAE Decode (Remote) and that's it.
