☁️BizyAir VAE Encode
ComfyUI Node Guide
- pixels
- vae
- LATENT
Every img2img-style workflow needs this step somewhere: you have real pixels and you need them in latent space before a sampler can do anything with them. VAE Encode is that translation, run in the direction of compression - a real image goes in, a latent representation comes out, ready to hand to a KSampler for a denoise pass at whatever strength your workflow calls for.
There's nothing exotic about what this node does conceptually; it's the same job ComfyUI's core VAE Encode performs. What's different is where it runs and what it expects on its inputs.
Inputs and outputs
Two required inputs:
pixels- the source image, typeIMAGE.vae- the VAE to encode with, typeBIZYAIR_VAE, which has to come from BizyAir Load VAE upstream. This is the one detail that catches people: you can't feed this a VAE loaded through a stock ComfyUI VAE Loader, since the types don't match.
Output is LATENT, a standard type - this one connects fine to any sampler downstream, BizyAir or otherwise, since latents themselves aren't given a special BizyAir type.
Installing BizyAir
Through ComfyUI Manager: search "BizyAir," install. Or clone it manually:
cd ComfyUI/custom_nodes && git clone https://github.com/siliconflow/BizyAir.git
Restart ComfyUI, or run comfy node install bizyair via Comfy-CLI. Set your API key on first use through the README's "click to login" step - the encode itself happens on SiliconFlow's cloud rather than on your local GPU, so this node won't function without one.
Common issues
The most common blocker is the vae input specifically: if you're building a workflow that mixes BizyAir nodes with stock ComfyUI ones, remember that this needs a BIZYAIR_VAE, which only BizyAir Load VAE produces. A regular VAE loader's output simply won't connect here.
Beyond that, the usual VAE-mismatch symptoms apply if your downstream results look off - encoding pixels with a VAE that doesn't match the model family you're about to sample with (say, an SDXL VAE feeding a Flux-based workflow) will produce a technically valid latent that decodes back to something wrong. Double-check the VAE loaded upstream actually matches your checkpoint before troubleshooting anything further down the chain.
If the node produces no output at all, check your API key before assuming anything about the image or VAE is the problem - a missing or invalid key is by far the most common reason any BizyAir node fails silently.
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| pixels | IMAGE | — | |
| vae | BIZYAIR_VAE | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| LATENT | LATENT | — |