n8n-nodes-sogni

Enhanced n8n Community Node for Sogni AI Image, Video, Audio, LLM & Creative-Workflow Generation

Generate AI images, videos, audio, LLM responses, and full hosted multi-step creative workflows on the Sogni AI Supernet — directly from your n8n workflows. Highlights:

• Image — text-to-image with full ControlNet (15 types), Qwen Image Edit with multi-reference context images, and a dynamic server-validated size-preset dropdown.
• Video — LTX-2.3, WAN 2.2, and Seedance families with cost estimation, ControlNet, image-to-video, sound-to-video, animate, and v2v workflows.
• Audio — generate music and instrumental audio (ACE-Step) with optional lyrics, BPM/time signature/key, composer mode, and cost estimation. (new in 1.7.0)
• LLM — Sogni Intelligence chat models with optional tool calling and vision input, a one-click toggle to expose Sogni's 24-tool hosted creative manifest to the model, and pre-flight cost estimates. (new in 1.7.0)
• Creative Workflow — start, list, fetch events for, and cancel hosted multi-step Sogni workflows (storyboard → keyframes → video, etc.) with optional poll-until-terminal mode. (new in 1.7.0)

This node pulls from your personal Sogni account—sign up for free to get 50 free Render credits per day. Under the hood, the project utilizes the @sogni-ai/sogni-intelligence-client (formerly @sogni-ai/sogni-client-wrapper), which is built on top of the official @sogni-ai/sogni-client SDK.

Features

Resources & Operations

Image Resource

• Generate: Create AI images with optional ControlNet guidance and a server-validated Size Preset dropdown that adapts to the chosen model + network.
• Edit: Edit images using Qwen Image Edit models with context images.

Video Resource

• Generate: Create AI videos with customizable parameters.
• Estimate Cost: Estimate token/USD cost before generation.

Audio Resource (new in 1.7.0)

• Generate: Create music or instrumental audio (ACE-Step) with optional lyrics, BPM, time signature, key/scale, composer mode, and creativity controls.
• Estimate Cost: Estimate token/USD cost for an audio request.

LLM Resource

• Generate: Create text responses with Sogni chat models. Supports custom tool calling via Tools JSON, and a one-click Enable Sogni Hosted Tools toggle that injects Sogni's 24-tool hosted creative manifest (generate_image, generate_video, generate_music, edit_image, animate_photo, apply_style, etc.) so the model can drive the platform end-to-end. (toggle new in 1.7.0)
• Estimate Cost: Pre-flight chat cost estimate for a model + messages + max_tokens combination. (new in 1.7.0)
• Get All: List all available Sogni LLM/chat models.

Creative Workflow Resource (new in 1.7.0)

• Start: Run a hosted multi-step workflow from a saved template (Template ID + Inputs JSON) or an inline plan (Inline Workflow JSON with steps[]). Optional Wait Until Terminal polling.
• Get: Fetch a workflow record by ID.
• List: List recent workflows (limit/offset).
• Get Events: Stream the event history for a workflow.
• Cancel: Cancel an in-flight workflow.

Model Resource

• Get All: List all available models.
• Get: Get specific model details.
• Get Most Popular: Get the model with the most active workers in a single call. (new in 1.7.0)

Account Resource

• Get Balance: Check SOGNI and Spark token balance.

See CHANGELOG.md for the full 1.7.0 release notes.

Installation

Option 1: Community Nodes UI (Recommended)

1. In n8n, open Settings ▸ Community Nodes
2. Select Install
3. Enter n8n-nodes-sogni
4. Confirm the installation (restart n8n if prompted)

Option 2: Manual Installation

# Run in your n8n installation directory
npm install n8n-nodes-sogni
# Restart your n8n instance after installation

Configuration

1. Add Credentials

1. In n8n, go to Credentials
2. Click Add Credential
3. Search for "Sogni AI"
4. Enter your credentials:
   • Username: Your Sogni account username
   • Password: Your Sogni account password
   • App ID: (Optional) Leave empty for auto-generation

2. Add Node to Workflow

1. Create or open a workflow
2. Click + to add a node
3. Search for "Sogni AI"
4. Select the node and configure

Basic Usage

💡 Tip: You can import example workflows directly into n8n! Create a new workflow, click the ⋮ (three dots) in the top right corner, select Import from File..., and choose a sample workflow from the ./examples folder.

Simple Image Generation

{
  "resource": "image",
  "operation": "generate",
  "modelId": "flux1-schnell-fp8",
  "positivePrompt": "A beautiful sunset over mountains",
  "network": "fast",
  "additionalFields": {
    "negativePrompt": "blurry, low quality",
    "steps": 20,
    "guidance": 7.5,
    "tokenType": "spark",
    "downloadImages": true
  }
}

ControlNet-Guided Generation

{
  "resource": "image",
  "operation": "generate",
  "modelId": "flux1-schnell-fp8",
  "positivePrompt": "A fantasy castle, magical, glowing",
  "network": "fast",
  "additionalFields": {
    "enableControlNet": true,
    "controlNetType": "canny",
    "controlNetImageProperty": "data",
    "controlNetStrength": 0.7,
    "controlNetMode": "balanced",
    "steps": 20,
    "downloadImages": true
  }
}

Video Generation

{
  "resource": "video",
  "operation": "generate",
  "videoModelId": "wan_v2.2-14b-fp8_t2v_lightx2v",
  "videoPositivePrompt": "A serene waterfall flowing through a lush green forest",
  "videoNetwork": "fast",
  "videoAdditionalFields": {
    "videoSettings": {
      "frames": 81,
      "fps": 16,
      "steps": 4,
      "guidance": 7.5
    },
    "output": {
      "downloadVideos": true,
      "outputFormat": "mp4",
      "width": 640,
      "height": 640
    },
    "advanced": {
      "tokenType": "spark",
      "timeout": 300000
    }
  }
}

Image Edit with Qwen

{
  "resource": "image",
  "operation": "edit",
  "imageEditModelId": "qwen_image_edit_2511_fp8_lightning",
  "imageEditPrompt": "Change the background to a beautiful sunset beach",
  "contextImage1Property": "data",
  "imageEditNetwork": "fast",
  "imageEditAdditionalFields": {
    "generationSettings": {
      "negativePrompt": "blurry, distorted",
      "numberOfMedia": 1
    },
    "output": {
      "downloadImages": true,
      "outputFormat": "png"
    },
    "advanced": {
      "tokenType": "spark"
    }
  }
}

ControlNet Types

All 15 ControlNet types are supported:

| Type | Description | Best For | |------|-------------|----------| | canny | Edge detection | Structure preservation | | scribble | Hand-drawn sketches | Sketch to image | | lineart | Line art extraction | Clean line drawings | | lineartanime | Anime line art | Anime/manga style | | softedge | Soft edge detection | Artistic control | | shuffle | Composition transfer | Layout preservation | | tile | Tiling patterns | Seamless textures | | inpaint | Masked area filling | Object removal/editing | | instrp2p | Instruction-based editing | Text-guided edits | | depth | Depth map | 3D structure | | normalbae | Normal map | Surface details | | openpose | Pose detection | Human pose transfer | | segmentation | Semantic segmentation | Layout control | | mlsd | Line segment detection | Architecture | | instantid | Identity preservation | Face consistency |

See ControlNet Guide for detailed usage instructions.

Parameters

Required Parameters

| Parameter | Type | Description | |-----------|------|-------------| | Model ID | string | AI model to use (e.g., flux1-schnell-fp8) | | Positive Prompt | string | What you want to generate | | Network | options | fast (SOGNI tokens) or relaxed (Spark tokens) |

Optional Parameters (Additional Fields)

| Parameter | Type | Default | Description | |-----------|------|---------|-------------| | Negative Prompt | string | "" | What to avoid | | Style Prompt | string | "" | Style description | | Number of Images | number | 1 | How many images (1-10) | | Steps | number | 20 | Inference steps (1-100) | | Guidance | number | 7.5 | Prompt adherence (0-30) | | Token Type | options | spark | spark or sogni | | Output Format | options | png | png or jpg | | Download Images | boolean | true | Download as binary data | | Size Preset | string | "" | Size preset ID | | Width | number | 1024 | Custom width (256-2048) | | Height | number | 1024 | Custom height (256-2048) | | Seed | number | random | Reproducibility seed | | Timeout | number | 600000 | Max wait time (ms) |

ControlNet Parameters

| Parameter | Type | Default | Description | |-----------|------|---------|-------------| | Enable ControlNet | boolean | false | Enable ControlNet | | ControlNet Type | options | canny | Type of ControlNet | | Control Image Property | string | data | Binary property name | | Strength | number | 0.5 | Control strength (0-1) | | Mode | options | balanced | balanced / prompt_priority / cn_priority | | Guidance Start | number | 0 | When to start (0-1) | | Guidance End | number | 1 | When to end (0-1) |

Video Generation Parameters

Required Parameters

| Parameter | Type | Description | |-----------|------|-------------| | Video Model ID | string | AI model to use for video generation | | Video Positive Prompt | string | What you want in the video | | Video Network | options | fast or relaxed |

Optional Video Parameters

| Parameter | Type | Default | Description | |-----------|------|---------|-------------| | Negative Prompt | string | "" | What to avoid in video | | Style Prompt | string | "" | Video style description | | Number of Videos | number | 1 | How many videos (1-4) | | Frames | number | 30 | Number of frames (10-120). For LTX-2 use 8n+1 frame counts | | Duration | number | auto | Optional seconds for model-aware frame calculation | | FPS | number | 30 | Frames per second (10-60) | | Steps | number | 20 | Inference steps (1-100) | | Guidance | number | 7.5 | Prompt adherence (0-30) | | Shift | number | model default | Optional motion intensity control | | TeaCache Threshold | number | model default | Optional T2V/I2V optimization control | | Sampler | string | model default | Optional sampler override | | Scheduler | string | model default | Optional scheduler override | | Reference Image Property | string | "" | Binary property for i2v/s2v/animate workflows | | Reference End Image Property | string | "" | Binary property for interpolation end frame | | Reference Audio Property | string | "" | Binary property for s2v workflows | | Reference Video Property | string | "" | Binary property for animate/v2v workflows | | Video Start | number | 0 | Optional source-video offset (seconds) | | Audio Start | number | 0 | Optional source-audio offset (seconds) | | Audio Duration | number | server default | Optional source-audio duration (seconds) | | Trim End Frame | boolean | false | Useful for transition stitching | | First Frame Strength | number | model default | LTX-2 keyframe interpolation control (0-1) | | Last Frame Strength | number | model default | LTX-2 keyframe interpolation control (0-1) | | SAM2 Coordinates (JSON) | string | "" | Animate-replace subject points, e.g. [{"x":0.5,"y":0.5}] | | Enable LTX-2 Video ControlNet | boolean | false | Enables controlNet for LTX v2v | | Video ControlNet Type | options | canny | canny, pose, depth, detailer | | Video ControlNet Strength | number | 0.8 | ControlNet strength for v2v | | Output Format | options | mp4 | Currently only mp4 is supported | | Download Videos | boolean | true | Download as binary data | | Width | number | 512 | Video width (256-1024) | | Height | number | 512 | Video height (256-1024) | | Timeout | number | auto | Max wait time (ms) | | Auto Resize Video Assets | boolean | true | Normalize/resize reference assets for video compatibility |

Image Edit Parameters (Qwen)

Required Parameters

| Parameter | Type | Description | |-----------|------|-------------| | Image Edit Model ID | string | Qwen Image Edit model to use | | Edit Prompt | string | Description of the edit to apply | | Context Image 1 | string | Binary property name for first context image (required) | | Network | options | fast or relaxed |

Optional Parameters

| Parameter | Type | Default | Description | |-----------|------|---------|-------------| | Context Image 2 | string | "" | Binary property for second context image | | Context Image 3 | string | "" | Binary property for third context image | | Negative Prompt | string | "" | What to avoid in result | | Style Prompt | string | "" | Style description | | Number of Images | number | 1 | How many images (1-10) | | Steps | number | auto | Inference steps (auto: 20 for standard, 4 for lightning) | | Guidance | number | auto | Prompt adherence (auto: 4.0 for standard, 1.0 for lightning) | | Download Images | boolean | true | Download as binary data | | Output Format | options | png | png or jpg | | Token Type | options | spark | spark or sogni | | Timeout | number | auto | Max wait time (ms) |

Qwen Image Edit Models

| Model ID | Description | Recommended Steps | |----------|-------------|-------------------| | qwen_image_edit_2511_fp8 | Standard quality model | 20 steps | | qwen_image_edit_2511_fp8_lightning | Fast lightning model | 4 steps |

Example Workflows

See the examples directory for complete workflow JSON files:

1. Basic Image Generation - Simple text-to-image
2. Batch Processing - Generate multiple images
3. Dynamic Model Selection - Auto-select best model
4. Scheduled Generation - Daily automated images
5. Video Generation - AI video creation with customizable parameters
6. Image Edit with Qwen - Edit images using context-aware Qwen models
7. Emotional Slothi Telegram Bot - Dynamic Qwen image-edit + Telegram posting
8. LTX-2 Video-to-Video ControlNet - Advanced v2v workflow with reference video + controls
9. WAN Animate-Replace with SAM2 - Subject-guided video replacement with reference image + source video
10. LTX-2 Text-to-Video - Minimal prompt-only LTX t2v workflow
11. LTX 2.3 Dynamic Text-to-Video - Auto-select an available ltx23-* model before generation
12. Sogni LLM Person Poem Page - Ask for a person's name in an n8n form, auto-select an available chat model, generate a witty rhyming poem, and show it on n8n's completion page
13. Sogni LLM Describe Uploaded Image - Upload an image in an n8n form, send it to the documented Qwen3.5 VLM path, and show the generated description on n8n's completion page

Output

Image Generation Output

JSON Output

{
  "projectId": "ABC123...",
  "modelId": "flux1-schnell-fp8",
  "prompt": "A beautiful sunset...",
  "imageUrls": [
    "https://complete-images-production.s3-accelerate.amazonaws.com/..."
  ],
  "completed": true,
  "jobs": [
    {
      "id": "JOB123...",
      "status": "completed"
    }
  ]
}

Binary Output (when downloadImages = true)

• image: First generated image
• image_1: Second image (if multiple)
• image_2: Third image (if multiple)
• etc.

Video Generation Output

JSON Output

{
  "projectId": "VID123...",
  "modelId": "video-model-id",
  "prompt": "A cat playing...",
  "videoUrls": [
    "https://complete-videos-production.s3-accelerate.amazonaws.com/..."
  ],
  "completed": true,
  "jobs": [
    {
      "id": "JOB456...",
      "status": "completed"
    }
  ]
}

Binary Output (when downloadVideos = true)

• video: First generated video
• video_1: Second video (if multiple)
• video_2: Third video (if multiple)
• etc.

Binary data includes:

• Proper MIME type (video/mp4)
• Filename: sogni_video_[projectId]_[index].[ext]
• Full resolution video data

Image Edit Output

JSON Output

{
  "projectId": "EDIT123...",
  "modelId": "qwen_image_edit_2511_fp8_lightning",
  "prompt": "Change the background to a sunset beach",
  "imageUrls": [
    "https://complete-images-production.s3-accelerate.amazonaws.com/..."
  ],
  "completed": true,
  "contextImagesCount": 1,
  "jobs": [
    {
      "id": "JOB789...",
      "status": "completed"
    }
  ]
}

Binary Output (when downloadImages = true)

• image: First edited image
• image_1: Second image (if multiple)
• image_2: Third image (if multiple)
• etc.

Binary data includes:

• Proper MIME type (image/png or image/jpeg)
• Filename: sogni_edit_[projectId]_[index].[ext]
• Full resolution edited image

Tips & Best Practices

1. Network Selection

• Fast Network:

  • Uses SOGNI tokens
  • Faster generation (seconds to minutes)
  • Higher cost
  • Best for: Time-sensitive applications

• Relaxed Network:

  • Uses Spark tokens
  • Slower generation (minutes to hours)
  • Lower cost
  • Best for: Batch processing, scheduled jobs

2. Model Selection

Popular models:

• flux1-schnell-fp8: Fast, high quality, 4 steps recommended
• coreml-sogni_artist_v1_768: Artistic style
• chroma-v.46-flash_fp8: Fast generation

Use "Get All Models" operation to see all available models.

3. Steps Configuration

• Flux models: 4-8 steps (optimized for speed)
• SD models: 15-30 steps (better quality)
• ControlNet: 20-30 steps (more control)

4. ControlNet Usage

• Start with strength 0.5 and adjust
• Use balanced mode for most cases
• Match ControlNet type to your control image
• See ControlNet Guide for details

5. Image Download

• Enable downloadImages to prevent URL expiry
• URLs expire after 24 hours
• Binary data is permanent in n8n
• Recommended for production workflows

6. Timeout Configuration

• Image - Fast network: 60,000ms (1 minute) usually enough
• Image - Relaxed network: 600,000ms (10 minutes) recommended
• Video - Fast network: 120,000ms (2 minutes) minimum
• Video - Relaxed network: 1,200,000ms (20 minutes) recommended
• Adjust based on complexity and model

7. Video Generation Tips

• Frame Count: Start with 30 frames for quick tests, increase for longer videos
• FPS: Use 30 fps for smooth motion, 10-15 fps for stylized/animated look
• Resolution: Start with 512x512 for faster generation, increase as needed
• Format: Currently only MP4 format is supported
• Models: Look for models with "video", "animation", or "motion" in their names

8. Image Edit Tips (Qwen)

• Model Selection: Use lightning variant for fast results (4 steps), standard for quality (20 steps)
• Context Images: Provide 1-3 reference images that inform the edit
• Edit Prompts: Be specific about what to change (e.g., "change background to beach" vs "make it better")
• Multiple References: Use 2-3 context images for complex edits like style transfer or object compositing
• Steps: Leave empty for auto-detection based on model, or override for fine control

Troubleshooting

"Insufficient funds" Error

Solution: Add more Spark or SOGNI tokens to your account

"Model not found" Error

Solution: Use "Get All Models" to see available models

"No binary data found" (ControlNet)

Solution:

1. Ensure previous node outputs binary data
2. Check the binary property name
3. Use "View" in n8n to inspect data

Workflow Times Out

Solution:

• Use relaxed network for slower but more reliable generation
• Increase timeout in Additional Fields
• Split large batches into smaller chunks

Images Not Downloaded

Solution:

• Check downloadImages is enabled
• Verify network connectivity
• Check n8n logs for download errors

"No binary data found" (Image Edit)

Solution:

1. Ensure previous node outputs binary data with the correct property name
2. Check contextImage1Property matches your binary property (default: data)
3. Use "View" in n8n to inspect binary data from previous node
4. For multiple context images, verify each property name is correct

Image Edit Results Unexpected

Solution:

• Use more specific edit prompts describing exactly what to change
• Try the standard model (qwen_image_edit_2511_fp8) for better quality
• Adjust guidance value (higher = more adherence to prompt)
• Provide additional context images for complex edits

Advanced Usage

Combining with Other Nodes

Discord Integration

Sogni Generate → HTTP Request (Discord Webhook)

Google Drive Storage

Sogni Generate → Google Drive (Upload File)

Social Media Posting

Sogni Generate → Twitter/Instagram API

Image Processing Pipeline

Load Image → Sogni ControlNet → Post-Processing → Save

Dynamic Prompts

Use expressions to generate dynamic prompts:

{{ "A " + $json.style + " image of " + $json.subject }}

Conditional ControlNet

Enable ControlNet based on conditions:

{{ $json.hasControlImage ? true : false }}

API Reference

Wrapper Library

This node uses the @sogni-ai/sogni-intelligence-client library. For standalone Node.js usage:

import { SogniClientWrapper } from '@sogni-ai/sogni-intelligence-client';

const client = new SogniClientWrapper({
  username: 'your-username',
  password: 'your-password',
  autoConnect: true,
});

const result = await client.createProject({
  modelId: 'flux1-schnell-fp8',
  positivePrompt: 'A beautiful sunset',
  network: 'fast',
  tokenType: 'spark',
  waitForCompletion: true,
});

See @sogni-ai/sogni-intelligence-client for full API documentation.

Version History

v1.5.7 (Current)

• 📦 Updated @sogni-ai/sogni-client-wrapper to v1.6.1
• 🔄 Pulled in wrapper-side upgrades from @sogni-ai/[email protected]
• ✅ Revalidated the n8n node against the latest wrapper release

v1.5.5

• 🤖 Added support for Sogni Intelligence with Sogni LLM models like Qwen3.5, including Generate and Get All chat model operations
• 🧠 Added advanced chat inputs:
  • Messages JSON
  • Tools JSON
  • Tool Choice JSON
• 🧩 Preserved incoming item JSON in LLM -> Generate outputs so looped/stateful workflows can carry execution state forward
• ⏱️ Extended chat-model lookup timeouts for more reliable LLM workflow startup
• 🧪 Added bundled LLM example workflows for:
  • person poem generation with n8n forms and completion pages
  • uploaded-image description using the documented Qwen3.5 VLM path
• 🖼️ Added a bundled sample upload image (examples/duck.jpg) for quick vision workflow testing
• 📚 Refreshed README and example docs around the current LLM and vision workflows
• 📦 Updated @sogni-ai/sogni-client-wrapper to v1.6.0

v1.4.2

• 🧪 Added dedicated LTX-2 text-to-video example workflow (examples/10-ltx2-text-to-video.json)
• 📦 Updated @sogni-ai/sogni-client-wrapper to v1.5.2
• 🎬 Added ltx23-* / ltx2.3-* video model detection and LTX frame normalization coverage
• 🧪 Added dynamic LTX 2.3 example workflow (examples/11-ltx23-dynamic-text-to-video.json)

v1.4.0

• 📦 Updated @sogni-ai/sogni-client-wrapper to v1.4.3
• 🎬 Added Video → Estimate Cost operation (wrapper estimateVideoCost)
• 🧠 Improved video model detection to include ltx2-*, ltx23-*, and wan_* model families
• 🧩 Added advanced video workflow inputs/controls for LTX/WAN (referenceVideo, referenceAudio, SAM2, keyframe strengths, video ControlNet)
• 🖼️ Aligned Qwen image-edit guidance defaults with wrapper (4.0 standard, 1.0 lightning)
• 🎥 Added Auto Resize Video Assets toggle for video generation

v1.3.1

• 📚 Enhanced README documentation for Image Edit feature
• 📝 Added Image Edit output section, tips, and troubleshooting

v1.3.0

• 🖼️ Added Qwen Image Edit support with multi-reference context images
• 📦 Updated @sogni-ai/sogni-client-wrapper to v1.4.0
• ⚡ Auto-detection of optimal steps based on model (20 for standard, 4 for lightning)
• 🎯 Up to 3 context images for sophisticated multi-reference editing

v1.2.0

• 🎬 Added full video generation support
• 📦 Updated @sogni-ai/sogni-client-wrapper to v1.2.0
• 🎥 MP4 video format support
• ⚙️ Configurable video parameters (frames, FPS, resolution)
• 📥 Automatic video download as binary data
• 🔍 Dedicated video model selection and filtering

v1.1.9

• 📝 Updated Sogni signup copy and highlighted ControlNet positioning

v1.1.8

• 🆕 Refreshed installation instructions and Sogni account links
• 📚 Added references to Sogni platform, docs, and SDK packages

v1.1.6

• ⚡ Changed default network from "relaxed" to "fast" for quicker generation
• 📝 Documentation updates

v1.1.5

• 🔧 Minor bug fixes and improvements
• 📝 Documentation updates

v1.1.0-1.1.4

• ✨ Added full ControlNet support (15 types)
• 📥 Added automatic image download
• 🔑 Enhanced appId auto-generation
• ⚙️ Improved default values
• 📚 Added ControlNet guide

v1.0.0

• Initial release
• Basic image generation
• Model and account operations
• Size presets support

Resources

• ControlNet Guide - Complete ControlNet usage guide
• Example Workflows - Ready-to-use workflow examples
• Sogni AI - Platform overview and product updates
• Sogni SDK Docs - Official SDK documentation
• Sogni Docs - Platform guides and API references
• Integration Guide - Complete integration guide

Support

For issues or questions:

1. Check this README
2. Review the ControlNet Guide
3. Check example workflows
4. Submit an issue on GitHub

License

MIT License - See LICENSE file for details

Credits

Built with:

• @sogni-ai/sogni-client - Official Sogni SDK
• @sogni-ai/sogni-intelligence-client - Enhanced wrapper library
• n8n - Workflow automation platform

Ready to generate and edit amazing AI images in your n8n workflows! 🎨✨