Nodes/ComfyUI Pollinations/Pollinations Free Image Generator
ComfyUI Node

Pollinations Free Image Generator

Free Images in ComfyUI With Zero VRAM, Zero API Keys

By ciga2011·Created 2 years ago·Updated 2 years ago· 5
Pollinations Free Image Generator
    • IMAGE
    prompta futuristic cityscape
    modelflux
    width900
    height600
    seed0
    nologotrue
    enhancetrue
    privatetrue

    The whole point: no model, no VRAM, no key

    Most custom nodes make your ComfyUI setup heavier - another checkpoint, another few gigabytes, another dependency. This one is the opposite. Pollinations Free Image Generator (the PollinationsNode) doesn't run any model locally at all. It just phones home to Pollinations' hosted image API and returns whatever comes back. No model download, no GPU pressure, no API key, no cost. If your machine chokes on Flux, this is the node that still gets you an image.

    The trade-off is the obvious one: you're at the mercy of someone else's servers. The API is genuinely free and genuinely unlimited-ish, which makes it great for quick ideation, placeholder art, and any workflow where "good enough right now" beats "best possible tonight." People lean on it for automated bulk generation in scripts - and that same popularity means it can be slow or flaky at peak times. Plan accordingly; it's a tool for velocity, not for your final hero shot.

    How it actually works

    Read the __init__.py and it's almost embarrassingly simple. The node builds a URL like https://image.pollinations.ai/prompt/{your_prompt}?model=flux&width=900&height=600&seed=0, fires a requests.get() at it, opens the response bytes with PIL, and converts the result into a standard ComfyUI IMAGE tensor. That's the whole pipeline. Your prompt goes straight into the URL path - which is worth remembering if you type anything with spaces or special characters you care about.

    Because there are no model files, the requirements.txt is empty. It only needs requests, PIL, torch, and numpy, all of which ship with a normal ComfyUI install. This is the rare custom node with zero dependency risk - no pip conflicts, nothing extra to babysit.

    The inputs that matter

    • prompt - your text prompt, a multiline box. The default ("a futuristic cityscape") is just the author's example, not a hint.
    • model - the dropdown with seven hosted options: flux, flux-realism, flux-anime, flux-3d, any-dark, flux-pro, and turbo. These are Pollinations' model names, not things you download. flux is the sensible default; flux-realism when you want photographic, turbo when you want it fast.
    • width / height - from 1 to 2048, defaulting to 900×600. Keep an eye on size: bigger images mean slower API round-trips.
    • seed - defaults to 0, up to 9999999999. Set it to get the same image back for a given prompt.

    The optional toggles - nologo, enhance, and private - all default to on. nologo drops the Pollinations watermark, enhance passes their upscaling-ish quality flag, and private keeps your prompt out of public use. The one output is a plain IMAGE tensor, so it plugs into anything that expects one - a Save Image node, a VAE decode, a detailer. That's the whole output story.

    Installing it

    The easy way: ComfyUI Manager → search "Pollinations" → install, then restart ComfyUI. Or manually:

    cd ComfyUI/custom_nodes
    git clone https://github.com/ciga2011/ComfyUI-Pollinations
    

    Restart ComfyUI and the node shows up under the Pollinations category in your node list. No extra dependencies to install, which is rare enough to be worth saying twice.

    Where people get burned

    • The API is the failure point. Most problems aren't your setup - they're Pollinations being slow, down, or rate-limited. A reddit thread literally titled "Looking for a free Pollinations AI alternative" exists because the service has gotten flaky under load. When the node throws, it's usually a non-200 status from upstream.
    • It can hang. The requests.get() call has no timeout, so if the API stalls, the node can sit there indefinitely instead of failing cleanly. If your workflow freezes mid-run, this node is the first suspect.
    • No offline use, obviously. This needs a live internet connection. Air-gapped machines need not apply.

    If you need deterministic, controllable output, you run Flux locally and accept the VRAM. If you need a free image in fifteen seconds and your GPU is busy, this is the fastest path in ComfyUI. It's not the node you build your reputation on - it's the node that unblocks you.

    CategoryPollinations

    Inputs (8)

    NameTypeDefaultDescription
    promptSTRINGa futuristic cityscape
    modelCOMBOflux7 options: flux, flux-realism, flux-anime, flux-3d, any-dark, flux-pro, +1
    widthINT9001–2048
    heightINT6001–2048
    seedINT00–9999999999
    nologooptBOOLEANtrue
    enhanceoptBOOLEANtrue
    privateoptBOOLEANtrue

    Outputs (1)

    NameTypeDescription
    IMAGEIMAGE