Skip to main content

​
What are Models in FlowScale?

Models in FlowScale AI are the powerful AI engines that drive your generative workflows. Just like in a local ComfyUI setup, these models need to be properly organized and accessible for your workflows to function. FlowScale simplifies this process by providing both a curated collection of public models and the ability to manage your own private model library.
Comprehensive models management dashboard showing public and private model libraries, usage statistics, and organization tools
For a deeper understanding of how models work in ComfyUI, refer to the official ComfyUI documentation on models.

​
Model Storage and Organization

​
Dedicated Model Storage

FlowScale provides each organization with a dedicated volume for model storage with:
  • Unlimited storage capacity - No restrictions on model library size
  • Organization-level access - All team members can access shared models
  • Persistent storage - Models remain available across all projects and sessions

​
ComfyUI-Compatible Folder Structure

FlowScale maintains the exact same folder structure as local ComfyUI installations, ensuring seamless compatibility: 
Models/
├── checkpoints/          # Stable Diffusion checkpoints
├── vae/                 # VAE models
├── loras/               # LoRA models
├── controlnet/          # ControlNet models
├── clip/                # CLIP models
├── unet/                # U-Net models
├── upscale_models/      # Super-resolution models
└── embeddings/          # Text embeddings
This structure ensures that models are automatically recognized by ComfyUI workflows without any additional configuration.

​
Accessing the Models Interface

You can access the same models interface from two convenient locations in FlowScale:

​
1. Global Models Tab

  • Located next to the Pods tab in the main navigation
  • Provides organization-wide model management
  • Access all models available to your organization

​
2. Project-Level Models Section

  • Available within each individual project for convenience
  • Shows the same organization-wide models as the global tab
  • Convenient access while working within a project context
  • No difference in functionality - just easier access during workflow development
Navigation interface highlighting the models tab location within the FlowScale application with clear breadcrumb and menu structure

​
Public Models Library

​
Curated Model Collection

FlowScale provides a comprehensive Public Models section featuring:
  • Latest AI models - Cutting-edge and trending models
  • Production-ready models - Tested and optimized for reliability
  • Popular model variants - Different versions of Stable Diffusion, FLUX, and more
  • Specialized models - ControlNet, LoRA, upscaling, and other task-specific models

​
Categories Available

  • Image Generation: Stable Diffusion (1.5, XL, 3.0), FLUX, Midjourney-style models
  • Style Transfer: LoRA models for specific art styles and concepts
  • Image Control: ControlNet models for pose, depth, edge guidance
  • Enhancement: Upscaling and restoration models
  • Text Processing: CLIP and embedding models
  • Video Generation: AnimateDiff models for image-to-video, text-to-video generation, motion transfer, and temporal consistency
The public library is continuously updated by the FlowScale team to include the latest and most effective models in the AI community.

​
Managing Private Models

​
When to Use Private Models

Use the Private Models section when:
  • The model you need isn’t available in the public library
  • You have custom-trained or fine-tuned models
  • You need proprietary models for your organization
  • You want to use specific model versions not in the public collection

​
Creating Folder Structure

Before uploading a model, ensure the correct folder exists:
  1. Check existing folders - See if the appropriate category folder exists
  2. Create new folder - If needed, create the folder matching ComfyUI conventions
  3. Follow naming conventions - Use standard ComfyUI folder names for compatibility
Dialog for creating organized model folders with type selection, naming conventions, and folder structure options

​
Uploading Models to FlowScale

​
Step-by-Step Upload Process

​
1. Navigate to the Upload Interface

  • Go to the Private Models section
  • Find or create the appropriate folder (e.g., checkpoints, loras)
  • Click the three-dot menu on the folder
  • Select Upload from the dropdown
Model upload interface showing multiple upload methods - direct upload, URL import, and search functionality with clear options

​
2. Choose Your Upload Method

The upload dialog provides three options: Option A: Direct URL Download
  • Paste the direct download URL of the model
  • FlowScale will download the model directly to your storage
  • Fastest method for publicly available models
Option B: Local File Upload
  • Upload model files from your computer
  • Drag and drop or browse to select files
  • Best for custom models or files you already have locally
Option C: Search and Import
  • Search by model name to find models from:
    • Civitai.com - Community-created models and LoRAs
    • Hugging Face - Research and open-source models
  • Browse search results and select the model you want
Detailed upload interface showing file selection, metadata extraction, progress tracking, and validation steps

​
3. Configure API Access (if needed)

For models from external platforms, you’ll need API credentials: For Civitai Models:
  • Get your API key from Civitai API Settings
  • Paste the API key in the upload dialog
  • Required for downloading from Civitai’s repository
For Hugging Face Models:
  • Get your access token from Hugging Face Token Settings
  • Paste the token in the upload dialog
  • Required for accessing Hugging Face models
Configuration interface for external service API keys required for model downloads and integrations

​
4. Start the Upload

  • Click Upload to begin the process
  • You’ll see a progress indicator
  • You can leave the page - the upload continues in the background

​
Upload Monitoring

Background Processing:
  • Models upload and process in the background
  • No need to stay on the models page during upload
  • Continue working on other tasks while models download
Completion Notifications:
  • Receive notification when model upload completes successfully
  • Get alerts if upload fails with error details
  • Check status anytime by clicking the bell icon in the top navigation to view notifications
  • All model download progress and completion status appears in the notifications panel
Real-time upload progress interface showing transfer status, validation steps, and completion indicators

​
Using Uploaded Models in ComfyUI

​
Accessing Models in Workflows

After uploading models to FlowScale:
  1. Open your ComfyUI workspace in any project
  2. Add model nodes to your workflow (Load Checkpoint, Load LoRA, etc.)
  3. Click the model dropdown in the node
  4. Select your uploaded model from the list

​
Refreshing Model List

If newly uploaded models don’t appear in ComfyUI:

​
Method 1: Refresh Button

  • Look for the Refresh button in the Actions Bar at the bottom of ComfyUI
  • Click to reload the model list
  • This updates ComfyUI’s model cache with newly uploaded models
Workspace refresh interface showing model synchronization and workspace update controls

​
Method 2: Restart ComfyUI

If refresh doesn’t work:
  1. Stop ComfyUI using the stop button
  2. Start ComfyUI again
  3. Models should now be visible in dropdowns
This fully reloads ComfyUI and ensures all models are properly detected.

​
Troubleshooting Model Issues

​
Model Not Visible in ComfyUI

Problem: Uploaded model doesn’t appear in node dropdowns in ComfyUI Solutions:
  1. Check folder placement - Ensure model is in the correct ComfyUI folder
  2. Verify folder spelling - Check that folder names match ComfyUI’s exact spelling requirements (e.g., checkpoints/, loras/, controlnet/)
  3. Refresh ComfyUI - Use the refresh button in the actions bar
  4. Restart ComfyUI - Stop and start ComfyUI to force reload
  5. Verify upload completion - Check that upload finished successfully
  6. Check file format - Ensure model file has correct extension (.safetensors, .ckpt, etc.)

​
Upload Failures

Problem: Model upload fails or gets stuck Solutions:
  1. Check internet connection - Ensure stable connection for large downloads
  2. Verify API keys - Ensure Civitai/Hugging Face credentials are correct
  3. Try different upload method - Switch between URL, local upload, or search

​
Getting Help

If you continue experiencing issues:
  1. Check error messages - Note any specific error codes or messages
  2. Try basic troubleshooting - Refresh, restart, re-upload
  3. Contact FlowScale support - Reach out via:
  4. Provide details - Include model name, upload method, and error messages
The FlowScale team can assist with model installation and troubleshooting to ensure your models work properly in your workflows.

​
Getting Started

Ready to start working with models? Check out our Import & Create Workflows guide to begin building your first AI-powered workflow using FlowScale’s model management system.
  • ComfyUI Workspace - Learn how to build workflows with models
  • Projects - Understanding project organization
  • Pods - Computing resources that run your models
⌘I