Skills API Reference

Machine-readable reference for all node types, their inputs, outputs, and capabilities. AI agents can crawl this page to understand which creative pipeline skills are available and how to compose them into workflows.

21 nodes|5 categories|Handle types: image, text, video, audio, 3d, easeCurve

Connection Rules

input

Image Input

imageInput

Upload or drag-and-drop an image to feed into the workflow. Supports PNG, JPG, WebP.

InputsNone
Outputs
  • imageThe loaded image as base64 data URL

Audio Input

audioInput

Upload an audio file (MP3, WAV, OGG) to feed into audio-capable nodes.

InputsNone
Outputs
  • audioAudio file as base64 data URL

Web Scraper

webScraper

Fetches a URL and extracts images and/or text from the page. Dual output.

Inputs
  • textURL to scrape (or type directly)
Outputs
  • imageExtracted images from the page
  • textCleaned page text content
Parameters
scrapeModeenumWhat to extract(default: all-images)
maxImagesnumberMax images to extract(default: 4)

processing

Prompt

prompt

Text input node. Type a prompt that feeds into generation or LLM nodes. Supports multi-segment prompts with + Add.

InputsNone
Outputs
  • textThe prompt text

Prompt Constructor

promptConstructor

Build prompts from a static template with dynamic variable slots filled by upstream text connections.

Inputs
  • textDynamic text values to insert into template slots
Outputs
  • textAssembled prompt
Parameters
staticTextstringTemplate text with {slot} placeholders

Combine Text

promptConcatenator

Joins multiple text inputs into a single output with a configurable separator.

Inputs
  • textText inputs to combine (accepts multiple)
Outputs
  • textCombined text
Parameters
separatorstringSeparator between joined texts(default: \n)

Image Iterator

imageIterator

Batches multiple images and runs the downstream workflow once per image. Supports random sampling.

Inputs
  • imageImages to iterate over (accepts multiple connections)
Outputs
  • imageCurrent image in iteration
Parameters
modeenum'all' or 'random'(default: all)
randomCountnumberHow many random images to pick(default: 1)
Runs all downstream nodes once per image. Supports local image uploads combined with connected inputs.

Text Iterator

textIterator

Splits incoming text by a separator and runs the downstream workflow once per segment.

Inputs
  • textText to split into segments
Outputs
  • textCurrent text segment in iteration
Parameters
splitModeenumHow to split: newline, period, hash, dash, custom(default: newline)
customSeparatorstringCustom split character (when splitMode is 'custom')
Each segment triggers a full downstream execution pass.

Array

arrayNode

Holds a list of text items. Each item becomes a separate text output during batch iteration. Can also receive items from upstream text connections.

Inputs
  • textIncoming text split by newlines and appended to local items
Outputs
  • textCurrent item during iteration
Works like an iterator: runs downstream once per item. Merges local items with connected upstream text (split by newlines).

List Selector

listSelector

Dropdown picker that selects one item from a configurable set of options and outputs it as text.

Inputs
  • textOptional: populates options from upstream
Outputs
  • textThe currently selected option
Parameters
selectedIndexnumberIndex of the selected option(default: 0)

Split Grid

splitGrid

Detects a grid layout in an image and splits it into individual cells. Useful for extracting frames from contact sheets.

Inputs
  • imageImage containing a grid layout
Outputs
  • imageIndividual grid cells as separate images

Annotate

annotation

Draw on an image using a Konva canvas — freehand, shapes, text overlays. Outputs the annotated image.

Inputs
  • imageBase image to draw on
Outputs
  • imageAnnotated image

Image Compare

imageCompare

Side-by-side comparison of two images with a draggable slider.

Inputs
  • imageTwo images to compare (accepts two connections)
OutputsNone

generation

Generate Image

nanoBanana

AI image generation using Gemini or Kie.ai models. Accepts prompt text and optional reference images. Supports multiple aspect ratios and model-specific parameters.

Inputs
  • imageReference image(s) for image-to-image generation
  • textText prompt describing what to generate
Outputs
  • imageGenerated image
Parameters
modelstringModel ID (e.g. nano-banana, nano-banana-pro, or Kie model)
aspectRatioenumOutput aspect ratio: 1:1, 16:9, 9:16, 4:3, 3:4
countnumberNumber of images to generate per run(default: 1)
Calls /api/generate. Supports Gemini native and 40+ Kie.ai models.

Generate Video

generateVideo

AI video generation from text prompts and/or reference images using Kie.ai video models (Sora, Veo, Kling, etc.).

Inputs
  • imageReference image for image-to-video generation
  • textText prompt describing the video
Outputs
  • videoGenerated video
Parameters
modelstringKie model ID
durationstringVideo duration in seconds
aspectRatioenumOutput aspect ratio

Generate 3D

generate3d

AI 3D model generation from images using Kie.ai 3D models.

Inputs
  • imageReference image for 3D reconstruction
Outputs
  • 3dGenerated 3D model (GLB)

LLM Generate

llmGenerate

Text generation using LLMs (Gemini, OpenAI, Anthropic, Groq). Can accept images for vision tasks. Outputs text.

Inputs
  • textInput prompt or context
  • imageImage(s) for vision/multimodal tasks
Outputs
  • textGenerated text response
Parameters
providerenumLLM provider: google, openai, anthropic, groq
modelstringModel ID within the provider
systemPromptstringSystem instructions for the LLM
Calls /api/llm. Supports streaming. Can chain with prompts and image inputs.

output

Output

output

Displays the final image result. End node of a workflow branch.

Inputs
  • imageImage to display
OutputsNone

Output Gallery

outputGallery

Collects and displays multiple images in a gallery grid. Useful with iterators.

Inputs
  • imageImages to collect (accepts multiple / iterated)
OutputsNone
Accumulates images across iterator runs into a single gallery view.

Video Stitch

videoStitch

Combines multiple video clips into a single output with transitions.

Inputs
  • videoVideo clips to stitch together
Outputs
  • videoCombined video

utility

Sticky Note

stickyNote

Colored note for workflow annotations. No data connections — purely visual.

InputsNone
OutputsNone