Ino Get Vae Download Config

The VAE is the easiest model component to get subtly wrong - the wrong autoencoder file means washed-out colors or straight-up noise, and the file you want lives in a repo with a name you'll never remember. Ino Get Vae Download Config is the pack's fix: a dropdown of known-good VAE models that outputs a complete download config as a JSON string, ready to feed the download nodes. No repo IDs, no subfolder guesswork.

It's part of the Ino Nodes pack from nobandegani, and it's the VAE member of a family that also covers CLIP, ControlNet, LoRA and image/video models - they all read from the pack's shipped CSV files and work identically.

How it works

The dropdown is populated from vae_files.csv, which ships inside the pack's data directory. Each row is one curated VAE with everything a downloader needs: the repo_id, filename, subfolder within the repo, the target model_type folder, and the weight_type (fp16, bf16...). Pick a model and the node serializes the row to JSON and hands it out of the single output, ModelConfig.

The current list is a solid real-world set: the Flux 1 VAE (ae.safetensors from black-forest-labs/FLUX.1-dev), the Wan 2.1 and 2.2 VAEs from Comfy-Org's repackaged repos, Kijai's bf16 Wan VAEs, and the Qwen-Image VAE, among others. That's exactly the "modern model" roster - VAE architectures have diverged a lot in the last couple of years, and matching the right one to your model is half the battle.

Feed ModelConfig into Ino Hugging Face Download Model and it fetches the file into the right place under ComfyUI/models/vae/ (or wherever the config's model_type points). Feed it to Ino Handle Download And Load Model and it goes further and loads it. The config string carries everything, so nothing is typed by hand.

Where it fits

Any workflow where you're assembling a model stack the "Ino way": grab a VAE config, a CLIP config, a diffusion config, and let the download/load nodes do the rest. It's especially handy for Flux and Wan setups, where the VAE is a separate download and getting the wrong one is a classic source of grey or noisy output.

The honest caveat: you're limited to what the CSV ships with. A niche or brand-new VAE won't be in the dropdown - for that you'd use the raw download node with manual fields. But for the mainstream models, it's the no-typo path, and it makes the "right VAE" a dropdown choice instead of a memory test.

Installing it

It ships with the whole Ino Nodes pack. ComfyUI Manager: search "ComfyUI Ino Nodes", install, restart. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/nobandegani/ComfyUI-InoNodes.git
cd comfyui_ino_nodes
pip install -r requirements.txt

Restart after. Pack-wide caveats apply: it's built on ComfyUI's V3 schema (README wants v0.18.1+), and requirements.txt installs inopyutils, the helper library every node imports.

Common issues

If the dropdown is stale or missing entries, it's loaded from the CSV at pack load time - restart ComfyUI after updating the pack. And the output is a JSON string meant for the pack's config-consuming download nodes, so don't paste it into a plain filename field. Search "Ino Get Vae Download Config" in the node menu (grouped under InoModelHelper).