comfyui-spicyapi Introduction
comfyui-spicyapi brings hosted SpicyAPI (https://spicyapi.ai) models into ComfyUI as nodes you can connect in a visual workflow. You can generate and edit images, create video and audio, work with text models, and pass the result of one step into the next.
The models run on SpicyAPI’s servers. You need an internet connection, a SpicyAPI API key, and a prepaid USD balance, but you do not need a GPU or local model downloads. Each model node shows a price for the options you select before you run it.
For example, you could generate an image, send it to an image-to-video node to add motion, and save the resulting clip. A workflow screenshot shows this kind of connection.
The original release snapshot lists 189 image, video, and audio nodes, plus 25 text models available through one Chat node. Your available nodes may differ because the extension rebuilds its model list from your account’s catalogue when ComfyUI starts. Browse the model catalogue (https://spicyapi.ai/models) or model IDs and starting prices for more detail.
How comfyui-spicyapi Works
Think of a ComfyUI workflow as a set of connected workstations. One node supplies an input, such as a prompt or image; a SpicyAPI node sends it to a hosted model; another node receives or saves the result. You can connect these steps to build a repeatable creative process.
A typical image-to-video workflow looks like this:
- Add Load Image and choose your starting image.
- Connect it to a video model’s image input and describe the motion you want.
- Connect the model’s video output to Save Video, then run the workflow.
Each model node presents the options that model accepts. Depending on the model, you may see choices such as duration, resolution, or audio. Image, video, and audio inputs appear as connections, called sockets. If a model accepts several references, connecting one can reveal another socket.
The outputs work like familiar ComfyUI data types: image nodes return IMAGE, video nodes return VIDEO, audio nodes return AUDIO, and text or transcription nodes return STRING. A video node can also return its last frame as an IMAGE when return_last_frame is enabled. Background removal provides an alpha mask, which you can use to separate a subject from its background.
When an option says model default, or a numeric option is set to -1, the extension leaves that choice to the hosted model. This is useful when you want to try a model before adjusting its settings.
comfyui-spicyapi Features
- Model-specific nodes: Find models through SpicyAPI > Video, Image, or Audio, or search the canvas by name. The controls on each node follow that model’s published options. For instance, changing a video node’s duration changes the requested clip and may change its price. The node-search screenshot shows how model descriptions and starting prices appear while browsing.
- Connected creative workflows: Pass an image into a video model, or connect a text result to an image model’s prompt. ComfyUI’s template browser includes examples for text-to-image, image-to-video, a Chat-written image prompt, text-to-speech, and LoRA.
- SpicyAPI Chat: Choose from the available text models, provide a prompt, and optionally add system instructions or images. Its
STRINGoutput can become another node’s prompt. - SpicyAPI LoRA: For model nodes that support LoRA, provide a direct, publicly accessible
.safetensorsdownload URL and a strength setting. A LoRA is an add-on that can influence a supported model’s output; changing its strength changes how strongly it is applied. You can chain LoRA nodes to use several together. See the LoRA workflow example. - Visible pricing and a spending limit: A node’s price badge updates as you change relevant options. Under Settings > SpicyAPI, Max cost per run stops a run before it starts if its quote exceeds your limit. The node reports the final charge afterward. See the settings screenshot.
- Control over repeat runs: Seeds start in
fixedmode, so queueing an unchanged graph reuses its first result. Set control after generate torandomizewhen you want a new result each run. For a model that does not accept a seed, the seed control governs whether the node runs again; it is not sent to that model. - Key kept out of shared workflows: The API key is stored in ComfyUI’s user folder or supplied through
SPICY_API_KEY, rather than being saved as a node setting in a workflow or image metadata.
Cost and cancellation: Requests are paid in USD from a prepaid balance. Rates vary by model and may be charged per image, second, character, or request. Check the node’s quote and the model catalogue (https://spicyapi.ai/models) before running. Cancelling a ComfyUI run stops your wait, but does not stop the hosted job; if that job succeeds, it is charged. Its result remains available in your SpicyAPI console for up to 14 days.
comfyui-spicyapi Models
The names below come from the release snapshot in the original README, not a guaranteed live list. Restart ComfyUI to refresh the nodes available to your account. Select a model by the task its node offers, such as text-to-image, image-to-video, speech, or editing, then check its node controls and linked model page for supported inputs and pricing. Similar names, including Spicy, Pro, Turbo, and LoRA, do not by themselves establish which option will look best or cost least for your project.
Video
Use video nodes to create clips or perform the specific video task named by a tool. For a first image-to-video experiment, connect a still image to a compatible node and save its VIDEO output.
Image
Choose an image-generation node to create artwork, an editing node when working from an existing image, or a named tool for a focused task such as removing a background or expanding an image. Nodes ending in LoRA support the LoRA workflow described above.
Audio
Audio nodes cover tasks such as speech generation, music generation, and transcription. Check the node’s inputs and output before connecting it: a transcription result is text (STRING), while generated sound is AUDIO.
Text
SpicyAPI Chat offers these 25 text models in the release snapshot: Claude Sonnet 5.5, Claude Opus 5.5, Grok 4.7, DeepSeek V4.1 Flash, GLM 5.3 Flash, GLM 5.3, DeepSeek V4 Pro, Gemini 3.7 Flash, Grok 4.6, DeepSeek V4 Flash, Claude Opus 5, Gemini 3.5 Flash-Lite, Kimi K3, Grok 4.5, Claude Sonnet 5, GLM 5.2, Claude Fable 5, Grok 4.3, Gemini 3.1 Pro Preview, Claude Sonnet 4.6, Claude Opus 4.6, Gemini 3 Flash Preview, Claude Opus 4.5, Claude Sonnet 4.5, and Gemini 2.5 Flash.
Use Chat when you want text you can read or pass into another node, such as a draft image prompt. The provided context does not establish a reliable artistic or performance comparison among these text models; check their available descriptions and prices when choosing one.
Troubleshooting comfyui-spicyapi
| What you see | What to do |
|---|---|
No SpicyAPI key is set | Open Settings > SpicyAPI, add your key, and try again. On a server, the key can instead be supplied through SPICY_API_KEY. |
40310 or an email-verification message | Open the verification link in your SpicyAPI sign-up email, then retry. |
40004 or “no deployment serves this combination” | Read the error for the option it names. Change that option on the model node, then run again. |
| A model on the website is missing from ComfyUI | Restart ComfyUI. Its model-node list is built when it starts; without a key, it uses the bundled release list. |
| A proxy or firewall interception message | Check whether HTTPS_PROXY is set and, if so, whether it points to a working proxy. A remote machine must be able to reach api.spicyapi.ai. |
An error with a request_id | Keep the ID with the error details. It identifies the specific request if you need help investigating it. |
Frequently asked questions
Can I use Seedance with a starting image?
Yes. Add a compatible Seedance image-to-video node, such as Seedance 2.5 · Image to Video, connect Load Image to its image input, write your prompt, and connect its video output to Save Video. Choose any available duration, resolution, or audio options on that node before running.
Do I need a powerful computer?
No GPU or downloaded model weights are required. The generation happens on SpicyAPI’s servers, so a laptop or CPU-only machine can use the nodes if it can run ComfyUI and connect to the service.
Why did an unchanged workflow return the same result?
Seeds begin in fixed mode. Set control after generate to randomize when you want a fresh result on each run.
Is a cancelled run free?
Not necessarily. Cancelling stops ComfyUI from waiting, but the hosted job continues and is charged if it succeeds. Check your SpicyAPI console for the result.
Will sharing my workflow expose my API key?
The key is not a node setting and is not included in shared workflow files or saved-image metadata. On a ComfyUI server that other people can access, the author recommends supplying the key through SPICY_API_KEY and setting SPICYAPI_LOCK_SETTINGS=1 so the key and spending limit cannot be changed over HTTP.
Can I use my own LoRA?
Yes, with a supported model node whose name ends in LoRA. Enter a direct .safetensors download URL that works without a login into SpicyAPI LoRA, choose its strength, and connect it to the model node’s loras input.
Learn More about comfyui-spicyapi
- ComfyUI setup and usage guide (https://docs.spicyapi.ai/docs/comfyui) for extension-specific guidance.
- SpicyAPI model catalogue (https://spicyapi.ai/models) and model IDs and starting prices for exploring choices before a run.
- SpicyAPI documentation (https://docs.spicyapi.ai) for further information about the service.
- Service status (https://status.spicyapi.ai) when you suspect a service interruption.
- Visual examples: connected image-to-video workflow, model search, settings, and LoRA chain.
