Nodes/ComfyUI-Pollinations-BYOP/🌸🎞️ Pollinations Video Gen URL (BYOP)
ComfyUI Node

🌸🎞️ Pollinations Video Gen URL (BYOP)

Turn a sentence into a video clip without touching your GPU β€” the Pollinations video node

By ChunkyPanda29Β·Created 6 months agoΒ·Updated about 15 hours agoΒ· 3
🌸🎞️ Pollinations Video Gen URL (BYOP)
  • image_input
  • video_url
β—„promptcinematic sunsetβ–Ί
β—„modelwanβ–Ί
β—„api_keyβ–Ί
β—„duration5β–Ί
β—„seed42β–Ί

Local video generation is a VRAM flex: Wan, Veo-class models, Seedance - the good stuff wants 16–24GB and a queue that could outlast your patience. 🌸🎞️ Pollinations Video Gen URL (BYOP) takes the other route. You type a prompt, pick a model, and it hands you a URL pointing at a clip generated on Pollinations' cloud. Your GPU does nothing. On an 8GB laptop that's the difference between "video gen isn't for me" and actually making a short clip between lunch bites.

It's one node in the ComfyUI-Pollinations-BYOP pack, the same pack that gives you image, audio, and text generation plus a login node. And yes - the display name ending in "URL" is the giveaway for how it behaves, so let's get that out of the way up front.

How it works

The node doesn't render video. It constructs a URL:

https://gen.pollinations.ai/video/<url-encoded prompt>?model=wan&duration=5&seed=42&key=sk_...

and returns it as a STRING called video_url. The actual generation happens when something requests that URL - so wire the output into a downloader, or open it in a browser, or you'll be sitting there watching a string that never turns into a file. This is the #1 source of "is the node broken?" confusion, and it's working exactly as designed.

There's a genuinely nice trick hidden in the optional inputs, though: image-to-video. Plug an IMAGE into image_input and the node uploads it to media.pollinations.ai/upload and appends it as an &image= parameter. That's how you animate a still - the README's recommended chain is literally Text Gen β†’ Image Gen β†’ this node.

The inputs that matter

  • prompt - the multiline box. Describe motion, camera, and style; the model does the rest.
  • model - dropdown from the pack's models.json, default wan (Wan 2.x family). Most of the list is πŸ’Ž paid-only - veo, veo-1080p, seedance-2.0, seedance-pro, grok-video-pro, p-video-1080p/720p. nova-reel and wan are your free-tier workhorses.
  • api_key (optional) - per-node override; blank uses the pack-wide saved key.
  • image_input (optional) - the IMAGE tensor for img2vid described above.
  • duration - 2–15 seconds, default 5. Longer clips queue slower, so don't max it out on a whim.
  • seed - default 42; pin it for reproducible results, drop it to -1 for a random roll.

One output: video_url (STRING).

Installing it

Pure API node - nothing to download, no model files, no GPU. ComfyUI Manager (search "Pollinations BYOP"), or:

cd ComfyUI/custom_nodes
git clone https://github.com/ChunkyPanda29/ComfyUI-Pollinations-BYOP.git

Restart and you're live.

Gotchas

The anonymous tier is free but rate-limited, and video is the slowest endpoint in the pack - expect the URL fetch to take a while during busy hours, and timeouts are a real thing. Adding your free key (via the pack's BYOP Login node) fixes most of it. And a warning that catches people: the πŸ’Ž models are paid, and nothing tells you "you just spent Pollen" until you check your balance. Start with wan and nova-reel before you let yourself near Veo.

CategoryPollinations/Video

Inputs (6)

NameTypeDefaultDescription
promptSTRINGcinematic sunsetβ€”
modelCOMBOwan17 options: google/gemini-omni-1.1-flash πŸ’Ž, grok-imagine-video-1.5 πŸ’Ž, grok-video-pro πŸ’Ž, happyhorse-1.1 πŸ’Ž, minimax-h3 πŸ’Ž, nova-reel, +11
api_keyoptSTRINGβ€”
image_inputoptIMAGEβ€”
durationoptINT52–15β€”
seedoptINT42-1–2147483647β€”

Outputs (1)

NameTypeDescription
video_urlSTRINGβ€”