FAL Model Select (Image)

FAL Model Select is the pack's dropdown, and it's the least exciting node in the group on purpose. FAL Text2Image and Image2Image both take model_id as a plain string - you could type fal-ai/flux/dev straight into the socket and skip this node entirely. What the dropdown buys you is not fat-fingering an endpoint id into a request that bills you per image. That's a real service.

Think of it as documentation as much as a selector. It's a live list of every model the pack can reach, so it's the fastest way to answer "wait, what do I have access to here?"

How it works

There's no magic. The node holds a static map in the pack's schema code that pairs friendly names with fal endpoint ids - flux-dev becomes fal-ai/flux/dev, and so on. Pick one, and select() hands the mapped string to whatever you wire it into. It makes no API call, spends no money, and needs no key. You can even ignore it and hardcode the id if you're scripting; the id is what actually travels.

The eight choices

The dropdown covers the whole FLUX family that fal hosts:

• flux-dev - the local-workhorse tier you could run yourself; steady, cheap-ish.
• flux-schnell - the fast, step-distilled one (Apache 2.0); good for quick drafts.
• flux-pro-v1.1 and flux-pro-new - the API-only flagships; the whole reason to use this pack.
• flux-kontext-pro and flux-kontext-dev - instruction-based image editing; Dev is the open-weights tier, Pro the API tier.
• flux-kontext-pro-multi - Kontext that takes multiple reference images instead of one.
• flux-schnell-redux - image variation, not true img2img (that catches people out in the Image2Image node).

If you're new to the pack: start with flux-dev for text-to-image to keep the bill low, and jump to flux-pro-new when you want the good stuff. Reach for a Kontext tier when your workflow involves editing existing images.

What it wires into

The single model_id STRING output plugs into the model_id input on FAL Text2Image or FAL Image2Image. Nothing else needs it.

Install

It ships with the rest of the pack, so install once and you get all four nodes:

cd ComfyUI/custom_nodes
git clone https://github.com/Paddel87/ComfyUI_fal_image.git
pip install -r ComfyUI_fal_image/requirements.txt
export FAL_KEY="fal-..."

Restart ComfyUI and it lands under fal.ai/Image. ComfyUI Manager (search "ComfyUI_fal_image") works too; on the Windows portable build, run pip with python_embeded/python.exe.

Common issues

Honestly, there aren't many. The node itself can't fail - it's a dropdown that returns a string. The two mistakes people make are upstream of it: picking flux-schnell-redux when they meant a real img2img model, and forgetting that which models are available changes as fal's catalog does, so if a choice ever looks stale, update the pack. When in doubt, the endpoint id it outputs is the ground truth - you can paste it into the fal docs and confirm what you're actually calling.

For a small solo-authored pack, this is the part that's quietly well-made: a hard guard rail on a string input that's otherwise too easy to get wrong.