Complete Parameters Reference

StreamDiffusionTD has 167+ parameters organized across multiple pages. This comprehensive reference covers all parameters with their exact names, types, and descriptions.

Page Navigation

  • Settings 1 - Core generation controls, prompts, seeds, step schedule
  • Settings 2 - Stream controls, connection settings, performance
  • Settings 3 - Configurations, presets, TensorRT engines
  • V2V - Video-to-video temporal consistency settings
  • ControlNet - Image conditioning controls (local & Daydream)
  • Models - Model loading, LoRA, embeddings
  • Install - Installation and backend configuration
  • Callbacks - TouchDesigner callback functions
  • About - Version info and UI settings

Settings 1

Core generation parameters, prompts, and step scheduling.

Status (Status) op('StreamDiffusionTD').par.Status Str
Current streaming status display
Default: Not Streaming Read-only status
Start Stream (Startstream) op('StreamDiffusionTD').par.Startstream Pulse
Initialize and start the diffusion stream
Default: 0 Action: Pulse
Stop Server (Stopstream) op('StreamDiffusionTD').par.Stopstream Pulse
Stop the entire StreamDiffusion server
Default: 0 Action: Pulse
Feedback Safe (Feedbacksafe) op('StreamDiffusionTD').par.Feedbacksafe Toggle
Limit processing speed for feedback loop compatibility (v0.2.99)
Default: Off Options: On | Off

Prompts

Normalize Prompt Weights (Normpweights) op('StreamDiffusionTD').par.Normpweights Toggle
Automatically normalize all prompt weights to sum to 1.0
Default: On Options: On | Off
Total Weight (Totalpweights) op('StreamDiffusionTD').par.Totalpweights Float
Total weight for all prompt blocks combined
Default: 1.0 Range: 0.0+
Interpolation (Setinterpolation) op('StreamDiffusionTD').par.Setinterpolation Menu
Method for interpolating between prompt blocks
Default: average Menu options
Prompt Blocks (Promptdict) op('StreamDiffusionTD').par.Promptdict Sequence
Multiple prompts with individual weights for blending
Default: "default fancy banana" Sequence Block

Step Schedule (T Index List)

Step Schedule (Tindexblock) op('StreamDiffusionTD').par.Tindexblock Sequence
The Step Schedule is similar to 'denoise' in img2img. Each step has a slider (1-49) which determines its position in the noise reduction sequence. Lower slider values initiate denoising sooner (more randomness), while higher values create images closer to the input. Adjustable while streaming.
Default: Model dependent Range: 1-49 per step

Seeds

Seed Noise Multiplier (Noisemult) op('StreamDiffusionTD').par.Noisemult Float
Multiplier for seed noise strength
Default: 1.0 Range: 0.0+
Seed Blocks (Seeddict) op('StreamDiffusionTD').par.Seeddict Sequence
Multiple seeds with individual weights for blending
Default: Random seed Sequence Block

Core Settings

CFG Type (Cfgtype) op('StreamDiffusionTD').par.Cfgtype Menu
The cfg_type for img2img mode
Default: self Menu options
Guidance Scale (Guidancescale) op('StreamDiffusionTD').par.Guidancescale Float
The CFG scale controlling adherence to prompt
Default: 1.0 Range: 0.0+
Delta (Delta) op('StreamDiffusionTD').par.Delta Float
The delta multiplier of virtual residual noise
Default: 1.0 Range: 0.0+
Main SD Model ID / Path (Modelid) op('StreamDiffusionTD').par.Modelid File
The name of the model to use for image generation
Default: stabilityai/sd-turbo HuggingFace ID or local path
My Models (Mymodels) op('StreamDiffusionTD').par.Mymodels Menu
Model list based on StreamDiffusion/streamdiffusionTD/working_models.json. Items automatically added after frame 200 of active streaming.
Default: stabilityai/sd-turbo Dynamic menu
Acceleration (Acceleration) op('StreamDiffusionTD').par.Acceleration Menu
The acceleration method (TensorRT, etc.)
Default: none Options: none | tensorrt
Width (Width) op('StreamDiffusionTD').par.Width Int
The width of the generated image
Default: 512 Range: 256-1024+
Height (Height) op('StreamDiffusionTD').par.Height Int
The height of the generated image
Default: 512 Range: 256-1024+

Settings 2

Stream control, connection settings, and performance parameters.

Status (Status2) op('StreamDiffusionTD').par.Status2 Str
Current streaming status display (page 2)
Default: Not Streaming Read-only status
Stream Active (Streamactive) op('StreamDiffusionTD').par.Streamactive Toggle
Indicates if streaming is currently active
Default: Off Read-only status
Server Active (Serveractive) op('StreamDiffusionTD').par.Serveractive Toggle
Indicates if the StreamDiffusion server is active
Default: Off Read-only status
SD Mode (Sdmode) op('StreamDiffusionTD').par.Sdmode Menu
Switch between text-to-image and image-to-image generation modes
Default: img2img Options: txt2img | img2img
Pause Stream (Pausestream) op('StreamDiffusionTD').par.Pausestream Toggle
Pause/resume streaming without stopping the server
Default: Off Options: On | Off
Process Delay (frames) (Delayframes) op('StreamDiffusionTD').par.Delayframes Int
Number of frames to delay processing
Default: 1 Range: 1+
Process Frame (Processframe) op('StreamDiffusionTD').par.Processframe Pulse
Manually trigger frame processing
Default: 0 Action: Pulse
Unload Model (Unloadstream) op('StreamDiffusionTD').par.Unloadstream Pulse
Unload model to free VRAM while keeping server alive
Default: 0 Action: Pulse

Connection Settings

Stream Out Name (Streamoutname) op('StreamDiffusionTD').par.Streamoutname Str
Shared memory name for communication with StreamDiffusion
Default: StreamDiffusionTD1_512-512 String identifier
Stream Out Mode (Streamoutmode) op('StreamDiffusionTD').par.Streamoutmode Menu
Method for streaming output data
Default: shared_mem Menu options
OSCin Port (Oscinport) op('StreamDiffusionTD').par.Oscinport Int
OSC input port for receiving messages in TouchDesigner
Default: 6503 Range: 1024-65535
OSCout Port (Oscoutport) op('StreamDiffusionTD').par.Oscoutport Int
OSC output port for sending messages from TouchDesigner
Default: 7187 Range: 1024-65535

Stream Settings

DEBUG (CMD stays open) (Debugcmd) op('StreamDiffusionTD').par.Debugcmd Toggle
Keep command window open for debugging purposes
Default: Off Options: On | Off
Visible CMD Window (Visiblewindow) op('StreamDiffusionTD').par.Visiblewindow Toggle
Show/hide the command window
Default: On Options: On | Off
Limit FPS (Limitfps) op('StreamDiffusionTD').par.Limitfps Int
Maximum frames per second for generation
Default: 60 Range: 1-120
Similar Image Filter (Imagefilter) op('StreamDiffusionTD').par.Imagefilter Toggle
Whether to enable similar image filter or not
Default: Off Options: On | Off
Similarity Filter Threshold (Filterthresh) op('StreamDiffusionTD').par.Filterthresh Float
The threshold for similar image filter
Default: 0.99 Range: 0.0-1.0
Max Skip Frame (Maxskipframe) op('StreamDiffusionTD').par.Maxskipframe Int
The max skip frame for similar image filter
Default: 1 Range: 1+

Additional Settings (some are disabled)

GPU ID (Gpuid) op('StreamDiffusionTD').par.Gpuid Int
GPU device ID for multi-GPU systems
Default: 0 Range: 0+
Frame Buffer Size (Framesize) op('StreamDiffusionTD').par.Framesize Int
The frame buffer size for denoising batch
Default: 1 Range: 1+
Add Noise (Addnoise) op('StreamDiffusionTD').par.Addnoise Toggle
Whether to add noise for following denoising steps or not
Default: On Options: On | Off
Denoise Batch (Denoisebatch) op('StreamDiffusionTD').par.Denoisebatch Toggle
Whether to use denoising batch or not
Default: On Options: On | Off
Negative Prompt (Negprompt) op('StreamDiffusionTD').par.Negprompt Str
This does not work due to the limitations of the distilled model process. Update as of version 0.2.9: maybe now if you set CFG to full mode and turn up CFG past 1.0, it might work -- but the result is just burning the negative latent into the image inversed. Not ideal!
Default: negative prompts don't work well.. Text input

Op Create

Create Synced Component (Synccompcreate) op('StreamDiffusionTD').par.Synccompcreate Pulse
Create a component synchronized with the stream timeline
Default: 0 Action: Pulse

Settings 3

Advanced configurations, presets, and model management.

Status (Status3) op('StreamDiffusionTD').par.Status3 Str
Current streaming status display (page 3)
Default: Not Streaming Read-only status

Op Configurations / Reset

Save OP Config (Writeconfig) op('StreamDiffusionTD').par.Writeconfig Toggle
Save current operator configuration to file
Default: Off Options: On | Off
Load OP Config (Loadconfig) op('StreamDiffusionTD').par.Loadconfig Toggle
Load saved operator configuration from file
Default: Off Options: On | Off
Reset Op (Resetop) op('StreamDiffusionTD').par.Resetop Pulse
Reset operator to default settings
Default: 0 Action: Pulse

Model Preset Configs

Preset Name (Presetname) op('StreamDiffusionTD').par.Presetname Menu
Select from saved model configuration presets
Default: Saved Model Configs Will Show Here Dynamic menu
Last Preset (Lastpreset) op('StreamDiffusionTD').par.Lastpreset Str
Display name of the last loaded preset
Default: Empty Read-only
Load Preset (Loadpreset) op('StreamDiffusionTD').par.Loadpreset Pulse
Load saved model preset configuration
Default: 0 Action: Pulse
Save Preset (Savepreset) op('StreamDiffusionTD').par.Savepreset Pulse
Save current model configuration as preset
Default: 0 Action: Pulse

TensorRT Engines

Select Engine (Engine) op('StreamDiffusionTD').par.Engine Menu
Select from available TensorRT engines
Default: Empty Dynamic menu
Load Engine Parameters (Loadengine) op('StreamDiffusionTD').par.Loadengine Pulse
Load parameters from selected TensorRT engine
Default: 0 Action: Pulse
Update Engine List (Updateengines) op('StreamDiffusionTD').par.Updateengines Pulse
Refresh the list of available TensorRT engines
Default: 0 Action: Pulse

Hugging Face Model Downloader (Deprecated)

Download Model (Dlhfmodel) op('StreamDiffusionTD').par.Dlhfmodel Pulse
Download model from HuggingFace (deprecated - use direct ID in model parameters instead)
Default: 0 Action: Pulse
Huggingface ID (Dlhfmodelid) op('StreamDiffusionTD').par.Dlhfmodelid Str
HuggingFace model ID for downloading
Default: Empty Text input
Model Type (download to) (Dlhfmodeltype) op('StreamDiffusionTD').par.Dlhfmodeltype Menu
Select model type for download destination
Default: LoRA Menu options
Check Model Info (Gethfmodelinfo) op('StreamDiffusionTD').par.Gethfmodelinfo Pulse
Get information about HuggingFace model
Default: 0 Action: Pulse

V2V

Video-to-video temporal consistency parameters for smooth animations.

StreamV2V Cached Attention

Enable V2V (Enablev2v) op('StreamDiffusionTD').par.Enablev2v Toggle
Enable video-to-video temporal consistency features
Default: Off Not compatible with TensorRT
Cache Attn Active (Caactive) op('StreamDiffusionTD').par.Caactive Toggle
Cache Attn Active enables the model to store attention maps from past frames, reusing them for processing new frames. This helps maintain consistency and reduces flicker in the video.
Default: On Options: On | Off
Cache Interval (Cacacheintvl) op('StreamDiffusionTD').par.Cacacheintvl Int
Cache Interval sets how often the feature bank updates with new frames. Lower values mean less frequent updates, capturing fewer changes and reducing computational load.
Default: 1 Range: 1+
Cache Max Frames (Cacachemaxfr) op('StreamDiffusionTD').par.Cacachemaxfr Int
Cache Max Frames limits the number of frames stored in the feature bank. Higher values store more frames, enhancing temporal consistency but using more memory and computational resources.
Default: 3 Range: 1+
Use Feature Injection (Causefeatinj) op('StreamDiffusionTD').par.Causefeatinj Toggle
Use Feature Injection incorporates details from past frames into the current frame's processing. It helps maintain continuity and reduces flicker in the video.
Default: On Options: On | Off
Feature Injection Strength (Cafeatinjstr) op('StreamDiffusionTD').par.Cafeatinjstr Float
Feature Injection Strength controls how much influence past frame details have on the current frame. Higher values increase consistency but may introduce artifacts.
Default: 0.7 Range: 0.0-1.0
Feature Similarity Threshold (Cafeatsimthr) op('StreamDiffusionTD').par.Cafeatsimthr Float
Feature Similarity Threshold sets the minimum similarity score for features from past frames to be reused. The model compares features using cosine similarity, only reusing those above this threshold.
Default: 0.3 Range: 0.0-1.0
Use Tome Cache (Causetomecache) op('StreamDiffusionTD').par.Causetomecache Toggle
Use Tome Cache activates an optimized cache for storing features efficiently. It reduces memory usage and enhances real-time processing performance.
Default: On Options: On | Off
Tome Ratio (Catomeratio) op('StreamDiffusionTD').par.Catomeratio Float
Tome Ratio defines the proportion of the total cache allocated to the Tome Cache. Higher ratios allocate more space to the Tome Cache, optimizing for speed at the cost of memory.
Default: 0.5 Range: 0.0-1.0
Use Grid (Causegrid) op('StreamDiffusionTD').par.Causegrid Toggle
Use Grid activates a grid-based approach for organizing and processing features. It divides the frame into a grid, improving the matching and fusion of features from past frames.
Default: On Options: On | Off

ControlNet

Image conditioning controls for both local and Daydream modes.

Local ControlNet (Single)

ControlNet Active (Usecontrolnet) op('StreamDiffusionTD').par.Usecontrolnet Toggle
Controls whether ControlNet image conditioning will be used. Can be toggled while streaming, but must be ON when hitting Start Stream.
Default: Off Must enable before starting
ControlNet Weight (Cnweight) op('StreamDiffusionTD').par.Cnweight Float
Strength of ControlNet conditioning influence
Default: 0.7 Range: 0.0-2.0
ControlNet Model (Cnmodel) op('StreamDiffusionTD').par.Cnmodel Menu
Select ControlNet model from available options
Default: download_controlnet Dynamic menu
ControlNet Folder (Cnfolder) op('StreamDiffusionTD').par.Cnfolder Folder
Folder containing ControlNet model files
Default: \models\ControlNet Directory path
Select ControlNet File (Cnmodelselect) op('StreamDiffusionTD').par.Cnmodelselect File
Direct file path to ControlNet model
Default: Empty File path
Download ControlNet (Cndownload) op('StreamDiffusionTD').par.Cndownload Menu
Download ControlNet models for different types
Default: depth Options: depth | openpose | canny | etc.

Daydream ControlNet (Multiple)

ControlNet (DayDream) (Cn) op('StreamDiffusionTD').par.Cn Sequence
Multiple ControlNet configurations for Daydream cloud mode
Multiple models available Sequence Block

Daydream ControlNet Preprocessors

Open Pose Confidence (Confidencethreshold) op('StreamDiffusionTD').par.Confidencethreshold Float
Confidence threshold for OpenPose detection
Default: 0.5 Range: 0.0-1.0
Canny low_threshold (Lowthreshold) op('StreamDiffusionTD').par.Lowthreshold Float
Lower threshold for Canny edge detection
Default: 100.0 Range: 0.0+
Canny high_threshold (Highthreshold) op('StreamDiffusionTD').par.Highthreshold Float
Upper threshold for Canny edge detection
Default: 200.0 Range: 0.0+

Models

Model loading, LoRA, and embedding configuration.

Download / Loading Models

Open Model Folder (Uiviewmodels) op('StreamDiffusionTD').par.Uiviewmodels Pulse
Open the models folder in file explorer
Default: 0 Action: Pulse
Force Model Pipeline (Forcemodeltype) op('StreamDiffusionTD').par.Forcemodeltype Menu
Forces the model loading process to use a specific model type
Default: auto Options: auto | sd15 | sdxl
SD Models Folder (Sdmodelsfolder) op('StreamDiffusionTD').par.Sdmodelsfolder Folder
Folder containing Stable Diffusion model files
Default: /models/Model Directory path
Main SD Model ID / Path (Modelid2) op('StreamDiffusionTD').par.Modelid2 Folder
The name of the model to use for image generation
Default: stabilityai/sd-turbo HuggingFace ID or local path
My Models (Mymodels2) op('StreamDiffusionTD').par.Mymodels2 Menu
Model list based on StreamDiffusion/streamdiffusionTD/working_models.json. Items automatically added after frame 200 of active streaming.
Default: stabilityai/sd-turbo Dynamic menu
Acceleration (Acceleration2) op('StreamDiffusionTD').par.Acceleration2 Menu
The acceleration method (TensorRT, etc.)
Default: none Options: none | tensorrt

LCM LoRA

Skip LCM LoRA (Skiplcm) op('StreamDiffusionTD').par.Skiplcm Toggle
Skip loading LCM LoRA (useful for models with merged LCM weights)
Default: Off Options: On | Off
Use Custom LCM (Usecustomlcm) op('StreamDiffusionTD').par.Usecustomlcm Toggle
Use custom LCM model instead of default
Default: Off Options: On | Off
Custom LCM (Customlcm) op('StreamDiffusionTD').par.Customlcm Folder
Path to custom LCM model
Default: Empty Directory path

LoRA Models Settings

Use LoRA (Uselora) op('StreamDiffusionTD').par.Uselora Toggle
Enable LoRA model loading and blending
Default: Off Options: On | Off
LoRA Folder (Lorafolder) op('StreamDiffusionTD').par.Lorafolder Folder
Folder containing LoRA model files
Default: \models\LoRA Directory path
Open LoRA Folder (Uiviewlora) op('StreamDiffusionTD').par.Uiviewlora Pulse
Open the LoRA folder in file explorer
Default: 0 Action: Pulse
LoRA Loader (Loradictblock) op('StreamDiffusionTD').par.Loradictblock Sequence
Multiple LoRA models with individual weights and file paths
Default: None Sequence Block

Textual Inversion (Embeddings)

Use Textual Inversion (Usetextualinv) op('StreamDiffusionTD').par.Usetextualinv Toggle
Enable textual inversion embeddings
Default: Off Options: On | Off
Add Embedding File (Textualinvblock) op('StreamDiffusionTD').par.Textualinvblock Sequence
Textual inversion embedding files with custom tokens
Default: None Sequence Block

Install

Installation, backend configuration, and setup parameters.

Backend (Backend) op('StreamDiffusionTD').par.Backend Menu
Choose between local GPU processing or Daydream cloud
Default: local Options: local | daydream
DayDream Apikey (Apikey) op('StreamDiffusionTD').par.Apikey Str
API key for Daydream cloud backend
Default: ✓ Loaded API key string
View Installation Guide (Viewinstallguide) op('StreamDiffusionTD').par.Viewinstallguide Pulse
Provides detailed instructions for the installation and update process
Default: 0 Action: Pulse

StreamDiffusion Install Location

Base Folder (Basefolder) op('StreamDiffusionTD').par.Basefolder Folder
The location on your computer where StreamDiffusion is downloaded and installed. This only needs to be set once.
Default: Empty Directory path

Installation Steps

1. Download StreamDiffusion (Clonestreamdiffusion) op('StreamDiffusionTD').par.Clonestreamdiffusion Pulse
If the Basefolder is empty, this will open a menu to choose a location to download the StreamDiffusion repository.
Default: 0 Action: Pulse
2. Install (venv + all req) (Installstreamdiffusion) op('StreamDiffusionTD').par.Installstreamdiffusion Pulse
Creates virtual environment and installs all necessary Python libraries. Checks for correct Python and CUDA versions.
Default: 0 Action: Pulse
3. Install TensorRT (optional) (Installtensorrt) op('StreamDiffusionTD').par.Installtensorrt Pulse
Optional: Enables faster FPS with limitations (512x512 resolution). Requires building TensorRT engine on first launch. NVIDIA GPU only.
Default: 0 Action: Pulse

Other Install Options

No Cache Install (Nocacheinstall) op('StreamDiffusionTD').par.Nocacheinstall Toggle
Disable pip caching during installation. Try enabling this if you encounter dependency errors.
Default: Off Options: On | Off
Set Hugging Face Cache (Sethfcache) op('StreamDiffusionTD').par.Sethfcache Toggle
Set custom location for HuggingFace model cache
Default: Off Options: On | Off
Hugging Face Cache (Hfcache) op('StreamDiffusionTD').par.Hfcache Folder
Custom folder for HuggingFace model cache
Default: /models Directory path
Manual Activate VENV (Openvenv) op('StreamDiffusionTD').par.Openvenv Pulse
For manual pip install/uninstall, etc (not needed for install)
Default: 0 Action: Pulse

Callbacks

TouchDesigner callback function controls.

Callback Dat (Callbackdat) op('StreamDiffusionTD').par.Callbackdat DAT
Reference to DAT containing callback functions
Default: /base1/StreamDiffusionTD_callbacks2 DAT reference
Edit Callbacks (Editcallbacksscript) op('StreamDiffusionTD').par.Editcallbacksscript Pulse
Open callback script for editing
Default: 0 Action: Pulse
Create Callbacks (Createcallbacks) op('StreamDiffusionTD').par.Createcallbacks Pulse
Create new callback script DAT
Default: 0 Action: Pulse
onReceiveFrame (Onreceiveframe) op('StreamDiffusionTD').par.Onreceiveframe Toggle
Enable callback when new frame is received from StreamDiffusion
Default: On Options: On | Off
onStreamStart (Onstreamstart) op('StreamDiffusionTD').par.Onstreamstart Toggle
Enable callback when streaming starts
Default: On Options: On | Off
onStreamEnd (Onstreamend) op('StreamDiffusionTD').par.Onstreamend Toggle
Enable callback when streaming ends
Default: On Options: On | Off
onFrameReady (Onframeready) op('StreamDiffusionTD').par.Onframeready Toggle
Enable callback when frame is ready for processing
Default: On Options: On | Off
onImageChange (Onimagechange) op('StreamDiffusionTD').par.Onimagechange Toggle
Enable callback when input image changes
Default: On Options: On | Off
Textport Debug Callbacks (Debugcallbacks) op('StreamDiffusionTD').par.Debugcallbacks Menu
Debug callback execution in textport
Default: None Menu options

About

Version information and UI settings.

Version (Txversion) op('StreamDiffusionTD').par.Txversion Str
Current version of StreamDiffusionTD
Default: 0.2.99 Version string
Date (Txdate) op('StreamDiffusionTD').par.Txdate Str
Release date of current version
Default: 08/19/2025 Date string
About Me (Txopenaboutme) op('StreamDiffusionTD').par.Txopenaboutme Pulse
Open information about the developer
Default: 0 Action: Pulse
Show Logs Level (Showlogs) op('StreamDiffusionTD').par.Showlogs Menu
Set logging level for debugging
Default: All Logs Menu options
UI Icons (Uiicons) op('StreamDiffusionTD').par.Uiicons Toggle
Show status icons in operator display
Default: On Options: On | Off
Show Built In (Showbuiltin) op('StreamDiffusionTD').par.Showbuiltin Toggle
Show built-in TouchDesigner parameters
Default: Off Options: On | Off