VAEDecode_motorway_edition

Every image that comes out of a ComfyUI sampler is a latent until the VAE decodes it back into pixels. VAEDecode is the node at the end of that chain - the one that turns your "samples" into the IMAGE you can look at. This _motorway_edition clone is the same decoder, but instead of a LATENT wire and a VAE wire going in, you type two key strings and the node pulls both out of the agilly1989 Motorway's shared dict.

If the Motorway is new to you: it's a pipe system from the ComfyUI_agilly1989_motorway pack. Start with MotorwayStart, then every _motorway_edition node reads its inputs from and writes its outputs to a shared dict via short key strings, all threaded down one MOTORWAY connection. These nodes are auto-generated clones of ComfyUI's own registry, so the decode math is identical to stock.

What it does. It takes a latent (samples) and a VAE and decodes the latent to pixels. Nothing clever - this is the plain, workhorse decoder. On the Motorway, INPUT_samples_key (default samples) is where the latent lives in the dict, INPUT_vae_key (default vae) is where the VAE lives, and OUTPUT_IMAGE_key (default IMAGE) is where the decoded image is stored for whatever comes next.

The inputs that matter. MOTORWAY is required, so this node has to sit in an existing Motorway. The three key fields are the whole story. And here's the trap that catches everyone on their first Motorway graph: the VAE loader stores its output under VAE uppercase, but this node looks it up as vae lowercase by default. So if you loaded the VAE with VAELoader_motorway_edition and left its defaults alone, you must set INPUT_vae_key to VAE or the decode will fail with 'MotorwayClass' object has no attribute 'hash_' - the pack's "key not in the Motorway" error. Same story for MODEL vs model and CLIP vs clip elsewhere in the pack. Uppercase in, lowercase out is the default wiring, and it doesn't line up.

Common issues. Beyond the case mismatch, the usual VAE advice applies even on the Motorway: decode with the VAE that matches your checkpoint. Swapping in a wrong-architecture VAE gives you flat color or noise, not a subtle color shift - the VAE is the codec, and the codec has to match the model that trained against it. Also, if the node isn't in your menu at all, the pack's current main branch has the clone registration commented out in __init__.py - re-enable it or use a version that has it on.

Install via ComfyUI Manager (search "ComfyUI_agilly1989_motorway") or git clone https://github.com/agilly1989/ComfyUI_agilly1989_motorway into custom_nodes, then restart. No models to download, no dependencies. Solo-dev beta pack - "if things break it's because I broke it" is literally the README, so keep the Motorway as a tidy-up layer, not the load-bearing part of your workflow.