![]()
ComfyUI custom nodes for creating AI texture processing workflows that integrate seamlessly with the NVIDIA RTX Remix Toolkit.
[!NOTE]
The RTX Remix Toolkit Integration is currently a work in progress and is not yet ready for production use. Beta releases may be available for testing and feedback on the RTX Remix Toolkit GitHub repository. If not, please check back later or contact the NVIDIA RTX Remix team on the RTX Remix Showcase Discord server.
/ComfyUI/custom_nodes/ foldergit clone https://github.com/NVIDIA/ComfyUI-RTX_Remix.git comfyui-rtx_remixcomfyui-rtx_remix folder and run install.bat (Windows) or install.sh (Linux/Mac)v0.3.48 or newer. Nodes V3 support was introduced in v0.3.48, and this pack uses the V3 schema API.This node pack provides two distinct approaches for integrating ComfyUI with the RTX Remix Toolkit.
Choose the one that fits your needs.
Mixing approaches is also possible but not recommended.
Best for: Most users who want to use AI texture processing in the RTX Remix Toolkit
How it works: The RTX Remix Toolkit controls ComfyUI. ComfyUI handles only the AI processing part - it receives image paths and parameters from the RTX Remix Toolkit, processes them, and returns the output file paths.
Key benefits:
Load Image to test your workflow without the RTX Remix Toolkit running๐ Jump to Integration Workflow Guide
Best for: Advanced users building complex workflows that require programmatic control of the RTX Remix Toolkit
How it works: ComfyUI controls the RTX Remix Toolkit remotely. You build comprehensive workflows that include layer management, texture ingestion, edit target selection, and asset manipulation - all controlled from within ComfyUI.
Key benefits:
Requirements:
๐ Jump to REST API Nodes Reference
The integration workflow system allows the RTX Remix Toolkit to execute your ComfyUI workflows without manual intervention. This is the recommended approach for most users.
The integration workflow system consists of three key steps:
Optionally, use Presets to save and compare different parameter configurations for your tagged inputs during workflow development.
Once exported, your workflow can be selected and run directly from within the RTX Remix Toolkit, which will:
Important: In this approach, ComfyUI only performs the AI processing. The RTX Remix Toolkit handles all project management, asset selection, layer organization, and texture replacement.
Create your ComfyUI workflow as you normally would. You can use any nodes available in ComfyUI for loading, processing, and generating textures.
Recommended Node: It's highly recommended to use RTX Remix Save Texture nodes to save your processed textures. While not strictly required, this node provides important benefits:
Alternative: You can use standard ComfyUI save nodes (like Save Image). If you output multiple textures, you'll need to manually configure texture type metadata for each output in the export dialog (expand the metadata section for each output to set texture types like albedo, normal, roughness, etc.) to ensure the RTX Remix Toolkit correctly matches outputs to shader inputs.
Tip: Use standard ComfyUI nodes for image processing, model loading, and AI processing. The RTX Remix Toolkit integration works with any workflow structure.
Development workflow: Integration workflows are easy to develop and test. Since they use standard ComfyUI nodes (plus the RTX Remix pack for tagging), you can iterate quickly:
Load Image nodes to test with sample textures without the RTX Remix Toolkit runningTagging tells the RTX Remix Toolkit which parts of your workflow should receive input data and which nodes produce the final outputs.
Input slots are connection points on nodes that can receive data from the RTX Remix Toolkit when the workflow is executed.
To tag an input slot:
Batch tagging: Select multiple nodes, then tag a slot. The same slot will be tagged on all compatible selected nodes.
Supported input types:
Any type that can be serialized to a primitive is supported, including:
STRING - For text parameters (file paths, names, configuration values, prompts, etc.)INT - For integer parameters (resolution, iterations, seed values, etc.)FLOAT - For decimal parameters (scale factors, strength values, etc.)BOOLEAN - For true/false flags (enable/disable features)Output nodes are the final nodes in your workflow that produce the textures/meshes the RTX Remix Toolkit should collect.
Important: Output nodes must save one or more file(s) to disk (e.g., images, textures, meshes). The RTX Remix Toolkit retrieves the output file paths from ComfyUI's output history, so any node that saves files will work.
To tag an output node:
RTX Remix Save Texture, Save Image, or any other save nodeRecommended: Use RTX Remix Save Texture nodes as outputs. These nodes:
Alternative: You can tag other save nodes (like Save Image). If you output multiple textures, you'll need to manually configure texture type metadata for each output in the export dialog (expand the metadata section for each output to set texture types like albedo, normal, roughness, etc.) to ensure the RTX Remix Toolkit correctly matches outputs to shader inputs.
Tip: Tagged inputs also appear in the Presets sidebar panel, where you can save different value configurations and organize inputs into groups.
Once you've tagged your inputs and outputs, export the workflow for use in the RTX Remix Toolkit.
To export:
RTX Remix Save Texture).json filename is derived from it), pick a Workflow Type, and optionally add a Description (used as the workflow tooltip in the RTX Remix Toolkit)The exported workflow is saved to the user/rtx-remix/workflows/ and user/rtx-remix/api_workflows/ directories in the ComfyUI installation and includes:
user/rtx-remix/api_workflows/) - A runtime-executable format enriched with RTX Remix metadata that the RTX Remix Toolkit uses to execute the workflowuser/rtx-remix/workflows/) - The complete workflow with all node configurations and tagged metadata - use this to open and edit the workflow in ComfyUIExport Dialog Features:
After exporting, your workflow becomes available in the RTX Remix Toolkit's workflow selection menu. When you run the workflow from the RTX Remix Toolkit, it will:
How data flows:
RTX Remix Save Texture or Save Image nodes)Presets let you save and switch between different sets of parameter values for your tagged inputs. They're useful for:
Presets only work with tagged inputs โ tag at least one input before creating presets.
Open the preset panel by clicking the โ (star) icon in the ComfyUI sidebar. The panel has two sections:
Presets use a default + override model:
Value resolution order: Custom preset override โ Default Values โ Current widget value
This keeps presets lightweight โ you only configure what's different, and everything else inherits the baseline.
| Action | How |
|---|---|
| Create | Click Create Preset in the panel header. Enter a name and optional description. The preset captures any values that currently differ from the Default Values. |
| Switch | Click a preset in the list to activate it. Its values are applied to all tagged widgets. If you have unsaved changes, you'll be prompted to save or discard. |
| Rename | Click the edit icon on a preset. The Default Values preset cannot be renamed. |
| Delete | Click the delete icon. Default Values can only be cleared when no other presets exist. |
Values can be edited in two ways โ both stay in sync:
Inputs whose values differ from the Default Values show a colored override indicator on the left side of the row, making it easy to see which values are customized in the current preset.
Per-input actions (click the โฎ menu on any input row):
Bulk actions (buttons above the inputs list):
Changes are held as pending until you click Save. You can also enable auto-save to skip the manual step.
Click the โพ dropdown arrow next to the Save button to access save options:
When a workflow has many tagged inputs, groups help keep them organized:
Groups are shared between the preset panel and the export dialog โ organizing in one place updates the other.
Download models directly into ComfyUI's models directory from HuggingFace, CivitAI, or a direct URL.
Purpose: Fetch model files during workflow setup or testing without leaving ComfyUI. Improves workflow portability by pairing template JSON workflows with ComfyUI Manager so missing node packs and models install without manual file browsing.
Dynamic inputs: This node is dynamic and will automatically show/hide inputs based on the detected source and options (e.g., enabling archive extraction or selecting HuggingFace/CivitAI/custom).
Inputs:
url (STRING) - Model URL (HuggingFace, CivitAI, or direct link)model_type - Target model folder (checkpoints, loras, vae, etc.)subdirectory - Optional subfolder within the model type directoryforce_download - Re-download even if the file existsextract_archive - Extract if the download is a zip/tar archivearchive_model_filename - Filename inside the archive (optional)extracted_model_hash - SHA256 hash of extracted model (optional)model_source - Auto-detected source (HuggingFace/CivitAI/custom)hf_repo_id - HuggingFace repository ID (auto-filled from URL)hf_filename - HuggingFace filename (auto-filled from URL)hf_token - HuggingFace token (optional, for gated models)civitai_model_id - CivitAI model version ID (auto-filled from URL)civitai_api_key - CivitAI API key (optional, for restricted downloads)custom_filename - Custom filename (auto-filled from URL)custom_hash - SHA256 hash for direct/custom downloads (auto-filled after download)Outputs:
model_name (ANY) - The saved model filenameCompatibility note:
The model_name output is ANY and is compatible with most model loading nodes. It connects directly to load node inputs and will download the model if missing, otherwise it uses the local file.
Features:
Category: RTX Remix
The recommended node for integration workflows. This node provides the cleanest integration experience, especially for workflows that output multiple texture types.
Purpose: Save processed textures to disk in a format and location that the RTX Remix Toolkit can use, with proper texture type metadata.
Inputs:
textures (IMAGE) - The texture image to savetexture_type - Type of texture (albedo, roughness, normal, metallic, etc.)file_extension - Format to save (.png or .exr)colorspace - Color space (linear or sRGB)Outputs:
file_path (STRING) - Path to the saved texture file(s)Features:
Category: RTX Remix
We provide example workflows demonstrating both integration and REST API approaches.
File: ./workflows/integration_pbrify.json
A complete example showing the integration workflow approach. This workflow:
RTX Remix Save Texture nodes with proper texture type metadata for each outputHow to use:
For advanced users building comprehensive workflows with full control of the RTX Remix Toolkit:
./workflows/restapi_pbrify.json - Full-featured REST API workflow with layer management./workflows/restapi_pbrify_lowVRAM.json - Memory-optimized version with project open/close nodesNote: These workflows use REST API nodes to control the RTX Remix Toolkit from within ComfyUI. They include layer creation, texture ingestion, and asset manipulation logic. This approach requires:
The REST API nodes enable ComfyUI to remotely control the RTX Remix Toolkit. Unlike integration workflows where the RTX Remix Toolkit controls ComfyUI, this approach puts ComfyUI in the driver's seat.
What's the difference?
When to use REST API nodes:
Requirements:
Development considerations: REST API workflows are harder to iterate on during development because:
Note: Most users should use Integration Workflows instead. REST API workflows are only recommended for advanced users building unique automation.
All REST API nodes are located under the RTX Remix > REST API category and prefixed with ๐ to indicate they communicate with the Toolkit via REST API.
JavaScript
45.8%
Python
27.3%
CSS
11.5%
HTML
8.7%
PowerShell
3.6%
Batchfile
2.0%
Shell
1.1%
![]()
ComfyUI custom nodes for creating AI texture processing workflows that integrate seamlessly with the NVIDIA RTX Remix Toolkit.
[!NOTE]
The RTX Remix Toolkit Integration is currently a work in progress and is not yet ready for production use. Beta releases may be available for testing and feedback on the RTX Remix Toolkit GitHub repository. If not, please check back later or contact the NVIDIA RTX Remix team on the RTX Remix Showcase Discord server.
/ComfyUI/custom_nodes/ foldergit clone https://github.com/NVIDIA/ComfyUI-RTX_Remix.git comfyui-rtx_remixcomfyui-rtx_remix folder and run install.bat (Windows) or install.sh (Linux/Mac)v0.3.48 or newer. Nodes V3 support was introduced in v0.3.48, and this pack uses the V3 schema API.This node pack provides two distinct approaches for integrating ComfyUI with the RTX Remix Toolkit.
Choose the one that fits your needs.
Mixing approaches is also possible but not recommended.
Best for: Most users who want to use AI texture processing in the RTX Remix Toolkit
How it works: The RTX Remix Toolkit controls ComfyUI. ComfyUI handles only the AI processing part - it receives image paths and parameters from the RTX Remix Toolkit, processes them, and returns the output file paths.
Key benefits:
Load Image to test your workflow without the RTX Remix Toolkit running๐ Jump to Integration Workflow Guide
Best for: Advanced users building complex workflows that require programmatic control of the RTX Remix Toolkit
How it works: ComfyUI controls the RTX Remix Toolkit remotely. You build comprehensive workflows that include layer management, texture ingestion, edit target selection, and asset manipulation - all controlled from within ComfyUI.
Key benefits:
Requirements:
๐ Jump to REST API Nodes Reference
The integration workflow system allows the RTX Remix Toolkit to execute your ComfyUI workflows without manual intervention. This is the recommended approach for most users.
The integration workflow system consists of three key steps:
Optionally, use Presets to save and compare different parameter configurations for your tagged inputs during workflow development.
Once exported, your workflow can be selected and run directly from within the RTX Remix Toolkit, which will:
Important: In this approach, ComfyUI only performs the AI processing. The RTX Remix Toolkit handles all project management, asset selection, layer organization, and texture replacement.
Create your ComfyUI workflow as you normally would. You can use any nodes available in ComfyUI for loading, processing, and generating textures.
Recommended Node: It's highly recommended to use RTX Remix Save Texture nodes to save your processed textures. While not strictly required, this node provides important benefits:
Alternative: You can use standard ComfyUI save nodes (like Save Image). If you output multiple textures, you'll need to manually configure texture type metadata for each output in the export dialog (expand the metadata section for each output to set texture types like albedo, normal, roughness, etc.) to ensure the RTX Remix Toolkit correctly matches outputs to shader inputs.
Tip: Use standard ComfyUI nodes for image processing, model loading, and AI processing. The RTX Remix Toolkit integration works with any workflow structure.
Development workflow: Integration workflows are easy to develop and test. Since they use standard ComfyUI nodes (plus the RTX Remix pack for tagging), you can iterate quickly:
Load Image nodes to test with sample textures without the RTX Remix Toolkit runningTagging tells the RTX Remix Toolkit which parts of your workflow should receive input data and which nodes produce the final outputs.
Input slots are connection points on nodes that can receive data from the RTX Remix Toolkit when the workflow is executed.
To tag an input slot:
Batch tagging: Select multiple nodes, then tag a slot. The same slot will be tagged on all compatible selected nodes.
Supported input types:
Any type that can be serialized to a primitive is supported, including:
STRING - For text parameters (file paths, names, configuration values, prompts, etc.)INT - For integer parameters (resolution, iterations, seed values, etc.)FLOAT - For decimal parameters (scale factors, strength values, etc.)BOOLEAN - For true/false flags (enable/disable features)Output nodes are the final nodes in your workflow that produce the textures/meshes the RTX Remix Toolkit should collect.
Important: Output nodes must save one or more file(s) to disk (e.g., images, textures, meshes). The RTX Remix Toolkit retrieves the output file paths from ComfyUI's output history, so any node that saves files will work.
To tag an output node:
RTX Remix Save Texture, Save Image, or any other save nodeRecommended: Use RTX Remix Save Texture nodes as outputs. These nodes:
Alternative: You can tag other save nodes (like Save Image). If you output multiple textures, you'll need to manually configure texture type metadata for each output in the export dialog (expand the metadata section for each output to set texture types like albedo, normal, roughness, etc.) to ensure the RTX Remix Toolkit correctly matches outputs to shader inputs.
Tip: Tagged inputs also appear in the Presets sidebar panel, where you can save different value configurations and organize inputs into groups.
Once you've tagged your inputs and outputs, export the workflow for use in the RTX Remix Toolkit.
To export:
RTX Remix Save Texture).json filename is derived from it), pick a Workflow Type, and optionally add a Description (used as the workflow tooltip in the RTX Remix Toolkit)The exported workflow is saved to the user/rtx-remix/workflows/ and user/rtx-remix/api_workflows/ directories in the ComfyUI installation and includes:
user/rtx-remix/api_workflows/) - A runtime-executable format enriched with RTX Remix metadata that the RTX Remix Toolkit uses to execute the workflowuser/rtx-remix/workflows/) - The complete workflow with all node configurations and tagged metadata - use this to open and edit the workflow in ComfyUIExport Dialog Features:
After exporting, your workflow becomes available in the RTX Remix Toolkit's workflow selection menu. When you run the workflow from the RTX Remix Toolkit, it will:
How data flows:
RTX Remix Save Texture or Save Image nodes)Presets let you save and switch between different sets of parameter values for your tagged inputs. They're useful for:
Presets only work with tagged inputs โ tag at least one input before creating presets.
Open the preset panel by clicking the โ (star) icon in the ComfyUI sidebar. The panel has two sections:
Presets use a default + override model:
Value resolution order: Custom preset override โ Default Values โ Current widget value
This keeps presets lightweight โ you only configure what's different, and everything else inherits the baseline.
| Action | How |
|---|---|
| Create | Click Create Preset in the panel header. Enter a name and optional description. The preset captures any values that currently differ from the Default Values. |
| Switch | Click a preset in the list to activate it. Its values are applied to all tagged widgets. If you have unsaved changes, you'll be prompted to save or discard. |
| Rename | Click the edit icon on a preset. The Default Values preset cannot be renamed. |
| Delete | Click the delete icon. Default Values can only be cleared when no other presets exist. |
Values can be edited in two ways โ both stay in sync:
Inputs whose values differ from the Default Values show a colored override indicator on the left side of the row, making it easy to see which values are customized in the current preset.
Per-input actions (click the โฎ menu on any input row):
Bulk actions (buttons above the inputs list):
Changes are held as pending until you click Save. You can also enable auto-save to skip the manual step.
Click the โพ dropdown arrow next to the Save button to access save options:
When a workflow has many tagged inputs, groups help keep them organized:
Groups are shared between the preset panel and the export dialog โ organizing in one place updates the other.
Download models directly into ComfyUI's models directory from HuggingFace, CivitAI, or a direct URL.
Purpose: Fetch model files during workflow setup or testing without leaving ComfyUI. Improves workflow portability by pairing template JSON workflows with ComfyUI Manager so missing node packs and models install without manual file browsing.
Dynamic inputs: This node is dynamic and will automatically show/hide inputs based on the detected source and options (e.g., enabling archive extraction or selecting HuggingFace/CivitAI/custom).
Inputs:
url (STRING) - Model URL (HuggingFace, CivitAI, or direct link)model_type - Target model folder (checkpoints, loras, vae, etc.)subdirectory - Optional subfolder within the model type directoryforce_download - Re-download even if the file existsextract_archive - Extract if the download is a zip/tar archivearchive_model_filename - Filename inside the archive (optional)extracted_model_hash - SHA256 hash of extracted model (optional)model_source - Auto-detected source (HuggingFace/CivitAI/custom)hf_repo_id - HuggingFace repository ID (auto-filled from URL)hf_filename - HuggingFace filename (auto-filled from URL)hf_token - HuggingFace token (optional, for gated models)civitai_model_id - CivitAI model version ID (auto-filled from URL)civitai_api_key - CivitAI API key (optional, for restricted downloads)custom_filename - Custom filename (auto-filled from URL)custom_hash - SHA256 hash for direct/custom downloads (auto-filled after download)Outputs:
model_name (ANY) - The saved model filenameCompatibility note:
The model_name output is ANY and is compatible with most model loading nodes. It connects directly to load node inputs and will download the model if missing, otherwise it uses the local file.
Features:
Category: RTX Remix
The recommended node for integration workflows. This node provides the cleanest integration experience, especially for workflows that output multiple texture types.
Purpose: Save processed textures to disk in a format and location that the RTX Remix Toolkit can use, with proper texture type metadata.
Inputs:
textures (IMAGE) - The texture image to savetexture_type - Type of texture (albedo, roughness, normal, metallic, etc.)file_extension - Format to save (.png or .exr)colorspace - Color space (linear or sRGB)Outputs:
file_path (STRING) - Path to the saved texture file(s)Features:
Category: RTX Remix
We provide example workflows demonstrating both integration and REST API approaches.
File: ./workflows/integration_pbrify.json
A complete example showing the integration workflow approach. This workflow:
RTX Remix Save Texture nodes with proper texture type metadata for each outputHow to use:
For advanced users building comprehensive workflows with full control of the RTX Remix Toolkit:
./workflows/restapi_pbrify.json - Full-featured REST API workflow with layer management./workflows/restapi_pbrify_lowVRAM.json - Memory-optimized version with project open/close nodesNote: These workflows use REST API nodes to control the RTX Remix Toolkit from within ComfyUI. They include layer creation, texture ingestion, and asset manipulation logic. This approach requires:
The REST API nodes enable ComfyUI to remotely control the RTX Remix Toolkit. Unlike integration workflows where the RTX Remix Toolkit controls ComfyUI, this approach puts ComfyUI in the driver's seat.
What's the difference?
When to use REST API nodes:
Requirements:
Development considerations: REST API workflows are harder to iterate on during development because:
Note: Most users should use Integration Workflows instead. REST API workflows are only recommended for advanced users building unique automation.
All REST API nodes are located under the RTX Remix > REST API category and prefixed with ๐ to indicate they communicate with the Toolkit via REST API.
JavaScript
45.8%
Python
27.3%
CSS
11.5%
HTML
8.7%
PowerShell
3.6%
Batchfile
2.0%
Shell
1.1%