- Image
- Generate
StreamDiffusion
Purpose
Section titled “Purpose”StreamDiffusion transforms incoming images according to a prompt and a selected StreamDiffusionTD configuration. That configuration determines the model, resolution and support for ControlNet or IP-Adapter inputs.
Place StreamDiffusion under Generate in the Image view. It requires a compatible StreamDiffusionTD installation, its Python environment, model assets and a working configuration.
On Voro’s Runtimes page, set StreamDiffusionTD Root and StreamDiffusion Python for that installation. Select the configuration in Config, connect a picture to in, and enter a Prompt.
Begin with a fixed Seed and feedback disabled so you can judge changes to the prompt.
Inputs and outputs
Section titled “Inputs and outputs”inis the image to transform.controlsupplies a ControlNet image when the configuration supports it and Control Image is enabled.stylesupplies an IP-Adapter image when configured and Style Image is on.outis the generated picture.
Controls
Section titled “Controls”- Config selects the pipeline configuration.
- Prompt and Prompt Blend steer the subject. Negative Prompt requires a pipeline using full classifier-free guidance.
- Seed Mode selects Fixed, Random or Blend. Seed Blend and Seed Interp control blended seeds.
- T-Index List selects diffusion timesteps within the configured pipeline.
- Feedback Loop and Feedback Strength feed generated images back into processing. Zoom, pan, rotation and Feedback Grade modify that loop.
- Engine Builder exposes model, resolution and optional-input choices. Build TRT Engine starts the build for those choices.
Build settings must match the model and optional inputs you intend to use. Building an engine can take time; it is separate from changing a live prompt. The builder uses the StreamDiffusionTD installation and model assets already configured on the machine.
Inputs
-
inimage required -
controlimage -
styleimage
Outputs
-
outimage
Parameters
Section titled “Parameters”Model
op('stream_diffusion').par.Configpath - Default:
td_config.yaml
Builder
op('stream_diffusion').par.Builder - Default:
Off- Options:
- Off, On
op('stream_diffusion').par.Build - Default:
None
op('stream_diffusion').par.Buildmodel - Default:
stabilityai/sdxl-turbo
op('stream_diffusion').par.Buildresolution - Default:
512x512
op('stream_diffusion').par.Buildtindex - Default:
"" (Empty String)
op('stream_diffusion').par.Buildcontrolnets - Default:
"" (Empty String)
op('stream_diffusion').par.Buildipamodel - Default:
"" (Empty String)
op('stream_diffusion').par.Buildipatokens - Default:
4- Range:
- 1 to 64
op('stream_diffusion').par.Buildipaencoder - Default:
"" (Empty String)
op('stream_diffusion').par.Buildbatch - Default:
1- Range:
- 1 to 4
op('stream_diffusion').par.Buildfolder - Default:
"" (Empty String)
Prompt
op('stream_diffusion').par.Prompt - Default:
"" (Empty String)
op('stream_diffusion').par.Promptblend - Default:
"" (Empty String)
op('stream_diffusion').par.Negprompt - Default:
"" (Empty String)
Generate
op('stream_diffusion').par.Tindex - Default:
"" (Empty String)
op('stream_diffusion').par.Delta - Default:
1- Range:
- 0 to 2
op('stream_diffusion').par.Seed - Default:
-1- Range:
- -1 to 2147483600
op('stream_diffusion').par.Seedblend - Default:
"" (Empty String)
op('stream_diffusion').par.Reset - Default:
None
Feedback
op('stream_diffusion').par.Feedback - Default:
Off- Options:
- Off, On
op('stream_diffusion').par.Feedbackstrength - Default:
0- Range:
- 0 to 1
op('stream_diffusion').par.Zoom - Default:
1- Range:
- 0.75 to 1.5
op('stream_diffusion').par.Panx - Default:
0- Range:
- -0.15 to 0.15
op('stream_diffusion').par.Pany - Default:
0- Range:
- -0.15 to 0.15
op('stream_diffusion').par.Rotation - Default:
0- Range:
- -10 to 10
op('stream_diffusion').par.Grade - Default:
Off- Options:
- Off, On
op('stream_diffusion').par.Gradestrength - Default:
0- Range:
- 0 to 1
op('stream_diffusion').par.Gradebrightness - Default:
0- Range:
- -1 to 1
op('stream_diffusion').par.Gradecontrast - Default:
1- Range:
- 0 to 2
op('stream_diffusion').par.Gradesaturation - Default:
1- Range:
- 0 to 2
op('stream_diffusion').par.Gradegamma - Default:
1- Range:
- 0 to 2
op('stream_diffusion').par.Gradeblacklevel - Default:
0- Range:
- 0 to 1
op('stream_diffusion').par.Gradehue - Default:
0- Range:
- -180 to 180
op('stream_diffusion').par.Gradetemperature - Default:
0- Range:
- -1 to 1
op('stream_diffusion').par.Gradeinvert - Default:
0- Range:
- 0 to 1
Control
op('stream_diffusion').par.Cnweight1 - Default:
-1- Range:
- -1 to 2
op('stream_diffusion').par.Cnweight2 - Default:
-1- Range:
- -1 to 2
op('stream_diffusion').par.Cnweight3 - Default:
-1- Range:
- -1 to 2
op('stream_diffusion').par.Usestyle - Default:
Off- Options:
- Off, On
op('stream_diffusion').par.Ipascale - Default:
-1- Range:
- -1 to 3