Skip to main content
ComfyUI-BudgetPixel is the official BudgetPixel node pack for ComfyUI. It adds one node per model — 58 nodes across image (FLUX 2, Seedream 5.0, Qwen-Image 3.0, Nano Banana, GPT-Image…), video (Kling v3, Seedance, Wan, MiniMax…), music and sound effects, upscaling, lip sync, and motion control — plus account utilities for live credits and pre-flight cost estimates. The nodes are thin clients for this API: each one submits a generation job, polls it, and downloads the result into your graph as a normal IMAGE, VIDEO, or AUDIO value. No local GPU or model downloads are needed for these nodes, and they compose freely with the rest of your local node graph.
The nodes use the developer API, which is included in every paid plan. Generations are billed in credits from your plan, identically to budgetpixel.com — each node’s help text shows its model’s price, and the BudgetPixel Cost Estimate node returns the exact discount-aware cost before you run a graph.

​
Install

  1. Open Manager → Custom Nodes Manager in ComfyUI.
  2. Search for BudgetPixel and click Install.
  3. Restart ComfyUI.

​
Add your API key

Create an API key (any paid plan), then provide it one of two ways:
Set BUDGETPIXEL_API_KEY in the environment that launches ComfyUI:
Your key is never stored in workflow JSON — the nodes deliberately have no key widget, so workflows you share or publish stay safe to share. Since other custom nodes run in the same ComfyUI process, we recommend a dedicated key for ComfyUI that you can revoke independently at any time.

​
Use the nodes

Everything lives under Add Node → BudgetPixel, one submenu per category: Generation runs on BudgetPixel’s side: nodes show progress, poll until the job succeeds, then pull the result into the graph. Local media inputs (frames, clips, audio) are pushed through temporary uploads automatically. Interrupting a ComfyUI run stops the wait, but a job that was already submitted keeps running — and billing — server-side. The pack ships example workflows (text-to-image, image-to-video, sound effects) that appear in ComfyUI’s template browser after install.
Video and audio nodes return ComfyUI’s core VIDEO/AUDIO types — use a 2025-or-newer ComfyUI.

​
How it maps to this API

Every node corresponds to one model endpoint — the node’s widgets are exactly that endpoint’s parameters, generated from this API’s OpenAPI spec. New models added to the API ship as new nodes in the next pack release. Under the hood the nodes follow the standard async job flow; everything documented there (statuses, output retention, rate limits) applies unchanged. Source, issues, and releases: github.com/budgetpixel-ai/comfyui-budgetpixel · Registry listing: registry.comfy.org