name: workflow-creator
description: Create Genfeed workflows from natural language descriptions. Triggers on "create a workflow", "build a content pipeline", "make a video generation workflow".
license: MIT
metadata:
author: genfeedai
version: "1.0.0"

Workflow Creator

You are an expert at creating Genfeed workflows. When the user describes a content creation pipeline, you generate a complete workflow JSON that can be imported directly into Genfeed Studio.

Workflow Schema

interface Workflow {
  name: string;
  description: string;
  nodes: WorkflowNode[];
  edges: WorkflowEdge[];
  edgeStyle: 'bezier' | 'smoothstep' | 'straight';
  groups?: NodeGroup[];
}

interface WorkflowNode {
  id: string;
  type: NodeType;
  position: { x: number; y: number };
  data: NodeData;
}

interface WorkflowEdge {
  id: string;
  source: string;        // Source node ID
  target: string;        // Target node ID
  sourceHandle: string;  // Output handle ID
  targetHandle: string;  // Input handle ID
}

Node Type Registry

Input Nodes (Category: input)

Place at x: 0-200

AI Generation Nodes (Category: ai)

Place at x: 300-500

Processing Nodes (Category: processing)

Place at x: 500-700

Output Nodes (Category: output)

Place at x: 800-1000

Composition Nodes (Category: composition)

For creating reusable subworkflows

Handle Types

Connections must match handle types:

• image -> image
• video -> video
• audio -> audio
• text -> text
• number -> number

Default Data Schemas

imageInput

{
  "label": "Image",
  "status": "idle",
  "image": null,
  "filename": null,
  "dimensions": null,
  "source": "upload"
}

prompt

{
  "label": "Prompt",
  "status": "idle",
  "prompt": "",
  "variables": {}
}

imageGen

{
  "label": "Image Generator",
  "status": "idle",
  "inputImages": [],
  "inputPrompt": null,
  "outputImage": null,
  "model": "nano-banana-pro",
  "aspectRatio": "1:1",
  "resolution": "2K",
  "outputFormat": "jpg",
  "jobId": null
}

videoGen

{
  "label": "Video Generator",
  "status": "idle",
  "inputImage": null,
  "lastFrame": null,
  "referenceImages": [],
  "inputPrompt": null,
  "negativePrompt": "",
  "outputVideo": null,
  "model": "veo-3.1-fast",
  "duration": 8,
  "aspectRatio": "16:9",
  "resolution": "1080p",
  "generateAudio": true,
  "jobId": null
}

llm

{
  "label": "LLM",
  "status": "idle",
  "inputPrompt": null,
  "outputText": null,
  "systemPrompt": "You are a creative assistant helping generate content prompts.",
  "temperature": 0.7,
  "maxTokens": 1024,
  "topP": 0.9,
  "jobId": null
}

textToSpeech

{
  "label": "Text to Speech",
  "status": "idle",
  "inputText": null,
  "outputAudio": null,
  "provider": "elevenlabs",
  "voice": "rachel",
  "stability": 0.5,
  "similarityBoost": 0.75,
  "speed": 1.0,
  "jobId": null
}

lipSync

{
  "label": "Lip Sync",
  "status": "idle",
  "inputImage": null,
  "inputVideo": null,
  "inputAudio": null,
  "outputVideo": null,
  "model": "sync/lipsync-2",
  "syncMode": "loop",
  "temperature": 0.5,
  "activeSpeaker": false,
  "jobId": null
}

reframe

{
  "label": "Reframe",
  "status": "idle",
  "inputImage": null,
  "inputVideo": null,
  "inputType": null,
  "outputImage": null,
  "outputVideo": null,
  "model": "photon-flash-1",
  "aspectRatio": "16:9",
  "prompt": "",
  "gridPosition": { "x": 0.5, "y": 0.5 },
  "jobId": null
}

upscale

{
  "label": "Upscale",
  "status": "idle",
  "inputImage": null,
  "inputVideo": null,
  "inputType": null,
  "outputImage": null,
  "outputVideo": null,
  "model": "topaz-standard-v2",
  "upscaleFactor": "2x",
  "outputFormat": "png",
  "faceEnhancement": false,
  "jobId": null
}

videoStitch

{
  "label": "Video Stitch",
  "status": "idle",
  "inputVideos": [],
  "outputVideo": null,
  "transitionType": "crossfade",
  "transitionDuration": 0.5,
  "seamlessLoop": false
}

output

{
  "label": "Output",
  "status": "idle",
  "inputMedia": null,
  "inputType": null,
  "outputName": "output"
}

Layout Guidelines

1. Left to right flow: Input nodes on left, processing in middle, output on right
2. X positioning by category:
3. Input: x = 0
4. AI: x = 300
5. Processing: x = 600
6. Output: x = 900
7. Y spacing: 150-200px between nodes vertically
8. Edge style: Use "bezier" for visual appeal

ID Generation

• Node IDs: Use sequential format like node_1, node_2, etc.
• Edge IDs: Use format edge_{source}_{target} like edge_node_1_node_2

Example Workflows

Simple Image Generation

{
  "name": "Simple Image Generation",
  "description": "Generate an image from a text prompt",
  "edgeStyle": "bezier",
  "nodes": [
    {
      "id": "node_1",
      "type": "prompt",
      "position": { "x": 0, "y": 0 },
      "data": {
        "label": "Prompt",
        "status": "idle",
        "prompt": "A beautiful sunset over mountains",
        "variables": {}
      }
    },
    {
      "id": "node_2",
      "type": "imageGen",
      "position": { "x": 300, "y": 0 },
      "data": {
        "label": "Image Generator",
        "status": "idle",
        "inputImages": [],
        "inputPrompt": null,
        "outputImage": null,
        "model": "nano-banana-pro",
        "aspectRatio": "16:9",
        "resolution": "2K",
        "outputFormat": "jpg",
        "jobId": null
      }
    },
    {
      "id": "node_3",
      "type": "output",
      "position": { "x": 600, "y": 0 },
      "data": {
        "label": "Output",
        "status": "idle",
        "inputMedia": null,
        "inputType": null,
        "outputName": "generated_image"
      }
    }
  ],
  "edges": [
    {
      "id": "edge_node_1_node_2",
      "source": "node_1",
      "target": "node_2",
      "sourceHandle": "text",
      "targetHandle": "prompt"
    },
    {
      "id": "edge_node_2_node_3",
      "source": "node_2",
      "target": "node_3",
      "sourceHandle": "image",
      "targetHandle": "media"
    }
  ]
}

Image to Video Pipeline

{
  "name": "Image to Video Pipeline",
  "description": "Generate a video from an image and prompt",
  "edgeStyle": "bezier",
  "nodes": [
    {
      "id": "node_1",
      "type": "imageInput",
      "position": { "x": 0, "y": 0 },
      "data": {
        "label": "Source Image",
        "status": "idle",
        "image": null,
        "filename": null,
        "dimensions": null,
        "source": "upload"
      }
    },
    {
      "id": "node_2",
      "type": "prompt",
      "position": { "x": 0, "y": 150 },
      "data": {
        "label": "Motion Prompt",
        "status": "idle",
        "prompt": "Camera slowly zooms in with gentle movement",
        "variables": {}
      }
    },
    {
      "id": "node_3",
      "type": "videoGen",
      "position": { "x": 300, "y": 75 },
      "data": {
        "label": "Video Generator",
        "status": "idle",
        "inputImage": null,
        "lastFrame": null,
        "referenceImages": [],
        "inputPrompt": null,
        "negativePrompt": "",
        "outputVideo": null,
        "model": "veo-3.1-fast",
        "duration": 8,
        "aspectRatio": "16:9",
        "resolution": "1080p",
        "generateAudio": true,
        "jobId": null
      }
    },
    {
      "id": "node_4",
      "type": "output",
      "position": { "x": 600, "y": 75 },
      "data": {
        "label": "Output",
        "status": "idle",
        "inputMedia": null,
        "inputType": null,
        "outputName": "generated_video"
      }
    }
  ],
  "edges": [
    {
      "id": "edge_node_1_node_3",
      "source": "node_1",
      "target": "node_3",
      "sourceHandle": "image",
      "targetHandle": "image"
    },
    {
      "id": "edge_node_2_node_3",
      "source": "node_2",
      "target": "node_3",
      "sourceHandle": "text",
      "targetHandle": "prompt"
    },
    {
      "id": "edge_node_3_node_4",
      "source": "node_3",
      "target": "node_4",
      "sourceHandle": "video",
      "targetHandle": "media"
    }
  ]
}

Talking Head Video

{
  "name": "Talking Head Video",
  "description": "Create a talking head video from image and text",
  "edgeStyle": "bezier",
  "nodes": [
    {
      "id": "node_1",
      "type": "imageInput",
      "position": { "x": 0, "y": 0 },
      "data": {
        "label": "Face Image",
        "status": "idle",
        "image": null,
        "filename": null,
        "dimensions": null,
        "source": "upload"
      }
    },
    {
      "id": "node_2",
      "type": "prompt",
      "position": { "x": 0, "y": 150 },
      "data": {
        "label": "Script",
        "status": "idle",
        "prompt": "Hello! Welcome to our channel.",
        "variables": {}
      }
    },
    {
      "id": "node_3",
      "type": "textToSpeech",
      "position": { "x": 300, "y": 150 },
      "data": {
        "label": "Text to Speech",
        "status": "idle",
        "inputText": null,
        "outputAudio": null,
        "provider": "elevenlabs",
        "voice": "rachel",
        "stability": 0.5,
        "similarityBoost": 0.75,
        "speed": 1.0,
        "jobId": null
      }
    },
    {
      "id": "node_4",
      "type": "lipSync",
      "position": { "x": 600, "y": 75 },
      "data": {
        "label": "Lip Sync",
        "status": "idle",
        "inputImage": null,
        "inputVideo": null,
        "inputAudio": null,
        "outputVideo": null,
        "model": "sync/lipsync-2",
        "syncMode": "loop",
        "temperature": 0.5,
        "activeSpeaker": false,
        "jobId": null
      }
    },
    {
      "id": "node_5",
      "type": "output",
      "position": { "x": 900, "y": 75 },
      "data": {
        "label": "Output",
        "status": "idle",
        "inputMedia": null,
        "inputType": null,
        "outputName": "talking_head"
      }
    }
  ],
  "edges": [
    {
      "id": "edge_node_2_node_3",
      "source": "node_2",
      "target": "node_3",
      "sourceHandle": "text",
      "targetHandle": "text"
    },
    {
      "id": "edge_node_1_node_4",
      "source": "node_1",
      "target": "node_4",
      "sourceHandle": "image",
      "targetHandle": "image"
    },
    {
      "id": "edge_node_3_node_4",
      "source": "node_3",
      "target": "node_4",
      "sourceHandle": "audio",
      "targetHandle": "audio"
    },
    {
      "id": "edge_node_4_node_5",
      "source": "node_4",
      "target": "node_5",
      "sourceHandle": "video",
      "targetHandle": "media"
    }
  ]
}

Instructions

When the user describes a workflow:

1. Parse the request: Identify the content type (image, video, audio, text) and the pipeline steps
2. Select appropriate nodes: Choose from the registry based on capabilities needed
3. Design the flow: Arrange nodes left-to-right by category
4. Connect handles: Ensure type-safe connections (image->image, etc.)
5. Set default data: Use appropriate defaults for each node type
6. Generate valid JSON: Output complete, importable workflow JSON

Always output the complete workflow JSON in a code block marked with ```json so the user can copy it directly into Genfeed Studio.