Z-Image Turbo VAE (CRT AutoDL)

Z-Image Turbo is the model that stole FLUX.2's launch: Alibaba's 6B, Apache 2.0, uncensored, and cheap enough to run on a consumer card. If you're running it in ComfyUI you need its VAE, and CRTAutoDLZImageTurboVAE is the loader that gets it for you. No file hunting, no curl from a repo you half-trust - it fetches the right ae.safetensors from Comfy-Org's official Z-Image split and hands you a VAE output.

Like the rest of the CRT AutoDL family, this node has no inputs. It's a pure download-on-first-use loader: on the first queue it pulls the VAE into ComfyUI/models/vae/ with a console progress bar, and on every run after that it just loads the file that's already there. Load it, wire the VAE output into your KSampler or VAE decode, done.

Why you'd reach for it

If you're new to the Z-Image stack, the VAE is the piece that looks trivially easy to get wrong. There are multiple Z-Image-flavored VAEs floating around - Qwen-Image VAEs, WAN VAEs, repacks with subtly different scaling - and a wrong one gives you washed-out or grey images that make you think the model is broken. This node pins the exact ae.safetensors Comfy-Org ships for Z-Image Turbo, which removes a whole failure class.

It also pairs naturally with the other AutoDL nodes: CRTAutoDLZImageTurboModel for the diffusion model and CRTAutoDLZImageTurboCLIP (a Lumina2-type encoder) complete the stack. That trio is arguably the fastest path from zero to a working local Z-Image setup in ComfyUI.

Gotchas

• First run isn't broken, it's downloading. The VAE is a couple of hundred MB, but on a slow connection it can look stalled. The CRT AutoDL progress line in your terminal is the source of truth.
• This VAE is Turbo-specific. Z-Image Turbo uses this particular autoencoder; if a workflow expects a different one (some edit pipelines do), don't force this file into it.
• No config to set. There's genuinely nothing to tune here. If you find yourself wanting options, you're probably overthinking it.

Install the pack via ComfyUI Manager (search CRT-Nodes) or git clone https://github.com/plugcrypt/CRT-Nodes.git into custom_nodes, install requirements.txt, restart, and queue once to trigger the download.