ComfyUI Node
⚗️🔬| Advanced VAE Decoder ^G2.1
An experimental alternative to the native ComfyUI VAEDecode node. It features options to bypass the standard output clamp so pixels can fall outside the [0.0, 1.0] range, and force tiled decoding to reduce VRAM usage on large images. ⚠️ Because this node is experimental, its parameters, behavior, or availability may change or be removed entirely without prior notice.
⚗️🔬| Advanced VAE Decoder ^G2.1
- samples
- vae
- IMAGE
◄allow_extended_range—►
◄low_vram_mode—►
Category⚡Z-Image/__experimental
Inputs (4)
| Name | Type | Default | Description |
|---|---|---|---|
| samples | LATENT | The latent representation to be decoded into an image. | |
| vae | VAE | The VAE model used to decode the latent input. | |
| allow_extended_range | BOOLEAN | When enabled, the standard numerical clamping on the VAE output is bypassed, permitting pixel values to exist outside the typical [0.0, 1.0] range, which could be beneficial for preserving high dynamic range data. Disable if image artifacts or distortion appear. | |
| low_vram_mode | BOOLEAN | When enabled, force the VAE decoding process to be split into smaller tiles. While this significantly reduces VRAM usage on large images, it results in slower processing speeds and a slight reduction in final image quality. This option should remain disabled unless you are operating under severe GPU memory constraints. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| IMAGE | IMAGE | The decoded image. |