comfyui-budgetpixel
Official BudgetPixel nodes for ComfyUI — 60+ hosted image, video, music & SFX models via your API key
Nodes (61)
When your image needs invented detail, not just bigger pixels
Put a price tag on a run before you spend
Your balance check, without leaving the canvas
Put your character's face on someone else's performance
The reliable workhorse of BudgetPixel's image shelf
BFL's next-gen image quality, minus the 24 GB of VRAM
BFL's sub-second model, now cheap enough to iterate on
The megapixel flagship when you need the BFL ceiling
Pay OpenAI's price for text rendering nothing local matches
XAI's model with the aesthetics reputation, at a flat rate
The do-everything graphics model for mockups and storyboards
Alibaba's character-consistent video, up to nine references
The fast Kling that ships its own audio
Hand your character a performance to copy
The two-frame classic with an optional soundtrack
The image model hiding inside the video maker's brand
The native-4K render that bills like a feature film
The 1080p flagship when Turbo isn't good enough
The 85-credit baseline that still feels like Kling
Nine-reference composition that goes to 4K
The one BudgetPixel node that does all four video jobs
Google's music model from a text prompt, and it even takes reference images
2K multimodal video with no license paperwork
The cheapest node that will actually sing your exact lyrics
The agentic image node that can search the web and run code before it draws
Full-length songs from a prompt, with a lyrics optimizer that writes for you
The newest MiniMax music model, lyrics included
Gemini 3.1 Flash Image, up to 4K, without a Google Cloud account
The flat-40-credit Google image node for when cost is the whole game
The 'thinking' Google image node for the assets that have to be right
Make a still photo speak, with actual body motion
The 10-credit, ~1-second image node you'll burn through in batches
Fast multi-image editing, and the prompt trick that makes it work
Cloud upscaling to 32 megapixels while your GPU does something else
The pack's bargain text-to-video, and its audio-driven trick
A talking head from one photo and a voice clip
The 15-credit node with the best text rendering in the pack
Unified generation and editing, with a negative prompt
The premium Qwen, when the image has to be believable
The newer, cheaper Qwen that's quietly the family's best default
Qwen Image 3.0 Pro in your graph — no VRAM required
SeeDance 1.5 Pro — the cheap way to animate a still
SeeDance 2.0 — four modes in one node, billed like it
SeeDance 2.0 Fast — the same Seedance 2.0, at half the resolution ceiling
SeeDance 2.0 Mini — the cost-performance pick of the Seedance line
SeeDance 2.5 — the flagship, where references become a creative medium
SeeDream 4.5 (BudgetPixel) — ByteDance's image model, minus the GPU
SeeDream 5.0 Lite — the cheap Seat of the Seedream line
SeeDream 5.0 Pro — the surgical edit, billed per input image
Sonilo Music — licensed instrumentals, exactly as long as you need
Sonilo SFX — foley on demand, one short prompt at a time
Sonilo Video Music — a soundtrack that actually follows your footage
Sonilo Video SFX — frame-accurate foley that watches the footage
Topaz Labs video upscale — pro-grade 4K60, billed by the second
BudgetPixel Upload Image — the boring utility node that makes the rest work
Wan 2.7 (BudgetPixel) — Alibaba's image model with a thinking mode
Wan 2.7 Pro — 4K text-to-image, but the edit mode tops out at 2K
Wan 2.7 (video) — Alibaba's video model, with a driving-audio track
Wan 3.0 — the all-in-one video node where input media is free
Wan 3.0 Prime — the faster tier of the all-in-one, for when patience costs more than credits
The hosted shortcut
comfyui-budgetpixel
Official BudgetPixel nodes for ComfyUI: generate with 60+ hosted image, video, music and sound-effect models — FLUX 2, Seedream 5.0, Qwen-Image 3.0, Kling v3, Nano Banana, GPT-Image, Seedance, Wan, and more — straight from your graph, using your BudgetPixel API key. No local GPU or model downloads needed for these nodes.
Status: beta — full API parity: 58 model nodes across image (26), video (17), music & sound effects (8), upscaling (3), lip sync (2) and motion control (2), plus account utilities. Everything the BudgetPixel API exposes is a node here.
Install
ComfyUI-Manager (recommended): search for BudgetPixel and install, then restart ComfyUI.
Manual:
cd ComfyUI/custom_nodes
git clone https://github.com/budgetpixel-ai/comfyui-budgetpixel.git
pip install -r comfyui-budgetpixel/requirements.txt
API key
The nodes call the BudgetPixel API, which is included in every paid plan. Create a key and provide it one of two ways:
- environment variable:
BUDGETPIXEL_API_KEY=bpx_live_... - a
budgetpixel.jsonfile inside thecomfyui-budgetpixelfolder:
{ "api_key": "bpx_live_..." }
Your key is never stored in workflow JSON — workflows you share stay safe to share. (That's also why the nodes have no api_key widget.)
Nodes
Everything lives under Add Node → BudgetPixel. One node per model — each node's help text shows its credit price, and every widget mirrors the API's typed parameters (enums, ranges, defaults).
- BudgetPixel/Image — text-to-image, image editing and multi-reference composition (
FLUX 2 Pro,Seedream 5.0 Pro,Nano Banana Pro, …). Image inputs are normalIMAGEconnections; batches map onto multi-reference models automatically. - BudgetPixel/Video — text-to-video, image-to-video (start/end frame), reference-to-video and video editing (
Kling v3,Seedance 2.5,Wan 3.0,MiniMax H3, …). Outputs are coreVIDEOvalues — wire into Save Video or frame extractors. - BudgetPixel/Audio — music (
Music 3.0,Lyria 3,Mureka V9, …) and sound effects (Sonilo SFX). Video-to-SFX and video-to-music take aVIDEOinput; Sonilo Video SFX also returns the source video with the effects mixed in. - BudgetPixel/Upscale — image upscalers plus Topaz Labs video upscaling.
- BudgetPixel/Lip Sync, BudgetPixel/Motion Control — talking-head sync from
IMAGE/VIDEO+AUDIO, and motion transfer from a reference video. - BudgetPixel/Account —
BudgetPixel Credits(live balance) andBudgetPixel Cost Estimate(exact pre-flight price, discount-aware for your account). - BudgetPixel/Media —
BudgetPixel Upload Image(temporary URL for an image, e.g. for debugging payloads).
Generation is asynchronous server-side: nodes show progress and poll until the result is ready, then download it into the graph. Interrupting a running graph stops the wait, but a job already submitted keeps running (and billing) server-side. Video and audio inputs are uploaded via the API's temporary storage; nothing is stored in your workflow file except normal node values.
VIDEOoutputs use the ComfyUI core video type — use a recent ComfyUI (2025+) for video and audio nodes.
Pricing
Generations are billed in BudgetPixel credits from your plan, identically to the web app — the node help text shows each model's price, and the Cost Estimate node gives an exact pre-flight number including your account's discounts. Prices summary: budgetpixel.com/pricing.
Development / how models stay current
Nodes are built at import time from committed schemas/*.json snapshots of the public OpenAPI spec — adding a model to the pack is a JSON diff, not new Python. Three ways they update:
- On demand: the
sync models from APIworkflow (Actions tab → run) regenerates the schemas from the live spec and opens a PR when models or params changed. Merge, bump the version inpyproject.toml, and the publish workflow ships to the Comfy Registry. - Locally:
python scripts/generate_schemas.py(addBUDGETPIXEL_API_KEYfor pricing text + availability filtering). - Verification:
python tests/test_nodes_build.py— offline check that every schema builds a well-formed node (no ComfyUI/torch needed); runs in CI on every PR.
scripts/e2e_smoke.py / scripts/e2e_node.py are live end-to-end tests against a real API (they spend credits; use a test account).
Issues and PRs welcome.