A simple but powerful utility node for ComfyUI that allows you to switch between the standard VAE Decode and the VAE Decode (Tiled) nodes without rewiring your workflow.
This node is designed to keep your graphs clean and make A/B testing different decoding methods quick and easy.
- Seamless Switching: A simple dropdown menu lets you select either the "default" or "tiled" VAE decoder on the fly.
- Dynamic UI: The node's interface is dynamic. When "default" is selected, the extra settings for the tiled decoder (like
tile_size,overlap, etc.) are completely hidden, keeping the node compact and clean. - Clean Workflows: Avoid messy rerouting nodes or having to manually disconnect and reconnect pipelines when you want to try a different decoder.
- Preserves Inputs: All standard inputs (
samples,vae) are maintained, and the node correctly passes the required parameters to the selected decoder.
- Navigate to your ComfyUI installation directory.
- Go into the
custom_nodesfolder. - Download or clone this repository into the
custom_nodesfolder. Ensure the final folder structure looks like this:ComfyUI/ └── custom_nodes/ └── VAE-Decode-Switch/ ├── __init__.py ├── vae_decode_switcher.py └── js/ └── vae_decode_switcher.js - Restart ComfyUI completely.
- After restarting ComfyUI, you can add the node by:
- Right-clicking on the graph and selecting
Add Node->Logic/Switchers->VAE Decode Switch. - Double-clicking the graph and searching for "VAE Decode Switch".
- Right-clicking on the graph and selecting
- Connect your
samples(LATENT) andvae(VAE) inputs as you would with a standard decoder. - Use the
select_decoderdropdown to choose your desired method.- default: Uses the standard
VAEDecode. The node will shrink to hide the other settings. - tiled: Uses the
VAEDecodeTiled. The node will expand to show the tiled-specific settings.
- default: Uses the standard
