spicyapiDocs
Main content

ComfyUI nodes

Run SpicyAPI's image, video and audio models (Seedance, Kling, Wan, MiniMax H3, Suno, GPT Image, Nano Banana and more) as ComfyUI nodes. Install the node pack, add your key, run your first node and keep every run's cost in view.

The SpicyAPI node pack for ComfyUI turns every SpicyAPI image, video and audio model into a ComfyUI node: more than 180 of them, plus a SpicyAPI Chat node for text models and a SpicyAPI LoRA node. The models run on SpicyAPI's servers, so you need an internet connection and an API key, not a graphics card or downloaded weights. Each node shows its price before it runs and prints what the run cost when it finishes.

The pack is free and open source (MIT) at github.com/spicyapi-ai/comfyui-spicyapi. It installs nothing beyond ComfyUI itself.

The short version

  1. In ComfyUI, click Manager, search for SpicyAPI, click Install and restart ComfyUI. Without Manager, run git clone https://github.com/spicyapi-ai/comfyui-spicyapi in ComfyUI/custom_nodes instead.
  2. Open Settings > SpicyAPI, paste your API key, and restart ComfyUI once more.
  3. Double-click the canvas, type a model name such as seedance or kling, connect a Save node to its output and press Run.

Is this the right tool?

You want to…Use
Chain models with other ComfyUI nodes in a visual graphThe ComfyUI nodes (this page)
Try a model once in your browser, with nothing installedThe Playground tab on any model at spicyapi.ai/explore
Generate from a terminal or a small scriptThe CLI
Let an AI assistant browse models and generate for youThe MCP server
Add generation to your own applicationThe SDKs or the HTTP API in the Quickstart

All of them use the same account, balance, models and prices.

Before you start

  • A recent ComfyUI. The nodes use ComfyUI's V3 node API and were tested with ComfyUI 0.39. Any install that loads folders from custom_nodes works: the desktop app, the portable build or a manual install.
  • An API key. Create one in the console; it starts with sk-spicy-. See Account setup if you do not have an account yet.
  • A balance. Every run is paid from your prepaid USD balance. See Billing.
  • Access to api.spicyapi.ai from the machine that runs ComfyUI. A proxy set through HTTPS_PROXY is honoured.

Install

Install the node pack

The quickest way is ComfyUI Manager: click Manager in the top bar, search for SpicyAPI and click Install. In the legacy Manager menu, the same search is under Custom Nodes Manager.

If the legacy menu does not find it, its node list is still cached from before the pack was added. Switch DB: Channel (1day cache) to DB: Channel (remote) and search again.

No Manager? Clone the pack yourself:

cd ComfyUI/custom_nodes
git clone https://github.com/spicyapi-ai/comfyui-spicyapi

No git? Download the ZIP and unpack it into ComfyUI/custom_nodes. There are no Python packages to install.

Restart ComfyUI

After the restart, the node menu has a SpicyAPI group with Image, Video, Audio, Text and Utils inside. To update later, click Update on the pack in Manager, or run git pull in its folder if you cloned it, then restart.

Add your API key

Open Settings > SpicyAPI > API key and paste your key. A notice confirms the key works and shows your available balance. Then restart ComfyUI once: the node list is built from your account's live catalogue when ComfyUI starts, and until a key is set the list bundled with the release is used.

The key is stored in a file in the ComfyUI user folder (user/spicyapi/config.json) that only your account can read. It is never written into a workflow file or into the metadata of an image ComfyUI saves, so sharing either one does not share your key. The settings field only shows the last four characters afterwards. To remove the key, clear the field.

On a server or in a container, set the SPICY_API_KEY environment variable instead. It takes priority over the settings field.

Run your first node

Add a model node

Double-click an empty spot on the canvas and type part of a model name, for example z-image, seedance, kling, wan, suno or nano banana. Node names match the model names on spicyapi.ai, such as Seedance 2.5 · Image to Video. You can also browse SpicyAPI > Image / Video / Audio in the node menu, grouped by maker.

Connect inputs and an output

Write a prompt on the node. If the model takes a picture, video or sound, connect a Load Image, Load Video or Load Audio node to the matching socket; the file is uploaded when the node runs. Connect Save Image, Save Video or Save Audio to the node's output, because ComfyUI only runs graphs that end in an output node.

Run it

Press Run. While the node works it shows the estimate and how long it has been running; when it finishes it shows the charge and the task ID. The task also appears under Logs in the console.

Prefer to start from a working graph? ComfyUI's template browser lists the pack's examples under comfyui-spicyapi: text to image, image to video, a chat model writing the prompt for an image model, text to speech, and LoRA.

Worked examples with their real results, prices and run times, plus the live list of uncensored nodes, are on the ComfyUI page at spicyapi.ai.

How the nodes work

Every node is generated from its model's published input schema, the same one the API validates requests against, so a node never offers an option the model would reject.

The model takes…On the node you get…
A fixed set of values, such as resolution or aspect ratioA dropdown
A number in a range, such as duration or strengthA slider or a number field
An on/off settingA switch
One image, video or audio fileAn input socket
Several reference filesA group of sockets that grows each time you connect one
A list of LoRA filesA loras socket, fed by SpicyAPI LoRA nodes
A structured list, such as a scripted dialogueA text field that takes JSON

A dropdown set to model default, or a number set to -1, is left out of the request so the model uses its own default.

Outputs are ordinary ComfyUI values, so any other node can take them:

  • Image models return IMAGE. Background removal also returns an alpha mask.
  • Video models return VIDEO, plus the clip's last frame as IMAGE when return_last_frame is on.
  • Audio models return AUDIO. A model that returns several clips, such as Suno, returns them as one batch and Save Audio writes one file each.
  • Transcription returns STRING.
  • Every node also returns its task_id, for the console logs or a support request.

Chat and LoRA nodes

SpicyAPI Chat sends a prompt, optional system instructions and optional images to any text model on SpicyAPI and returns the answer as a STRING. Wire it into an image or video node's prompt to have a language model write or refine prompts inside the graph. A model that does not read images refuses image input with nothing charged.

SpicyAPI LoRA adds one LoRA weights file, given as a direct download link to a .safetensors file, with a strength from 0 to 4. Chain several to stack them, and connect the last one to the loras input of a LoRA model node (their names end in LoRA). The link must open without a login.

What a run costs

  • The badge on each node shows the model's current rate for the options you picked, for example $0.0525/s for video or $0.01/image, and changes as you change them.
  • Every run is quoted first. Under Settings > SpicyAPI > Max cost per run, set a ceiling in US dollars. A run quoted above it stops before anything is charged.
  • The final charge appears on the node when the run finishes. Failed runs are not charged, except on models whose page says refusals are billed. See Billing.
  • Seeds start on fixed. Queueing the same graph twice reuses the first result instead of paying again. Switch control after generate to randomize to get a new result, and a new charge, on every run.
  • Cancelling stops the wait, not the job. A run you cancel in ComfyUI keeps going on SpicyAPI, is charged if it succeeds, and its result is in the console logs.

Results stay in your console for up to 14 days, or less if your account keeps them for a shorter time. Use ComfyUI's Save nodes to keep your own copy. See Retention.

Troubleshooting

You seeWhat to do
No SpicyAPI key is setAdd the key in Settings > SpicyAPI, or set SPICY_API_KEY, and run again.
SpicyAPI rejected this keyCopy the key again from the console; a revoked key stops working at once.
A model from the website is missingRestart ComfyUI. New models appear after the restart that follows their release.
40310Your account's email address is not verified. Open the link in your sign-up email.
40004No deployment serves that combination of options. Change the option named in the message.
This run could cost up to …The quote is above your Max cost per run. Raise the limit or pick cheaper options. Nothing was charged.
a proxy or firewall … may be intercepting the requestThe machine cannot reach api.spicyapi.ai cleanly. Check HTTPS_PROXY and firewall rules.

Errors from SpicyAPI end with a request_id or a task ID. Quote it to support and the request can be found exactly. Error codes are explained in Errors.

On this page