If you run into issues while installing tyDiffusion, check for solutions on the AI FAQ page, or contact support@tyflow.com for further assistance.
Enable DirectML support (AMD GPUs): enables DirectML support for AMD GPUs, as an alternative to PyTorch (NVIDIA GPUs).
Additional command line args: allows users to specify additional ComfyUI command line arguments, which will be processed when ComfyUI is restarted. A list of all supported arguments can be found here, or you can click the [>>] menu to select an argument from the list of presets.
Enable high-quality previews: when enabled, ComfyUI will display high-quality (TAESD) previews during generation, at a small performance cost. Disabling this setting will display low-quality (Latent2RGB) previews during generation.
Enable install/engine file logging: when enabled, tyDiffusion will export log files to disk while installing/running, which can be sent to support@tyflow.com if you need help figuring out why the tyDiffusion installer or ComfyUI engine are not functioning properly.
By default, tyDiffusion’s console windows are configured to close immediately after execution. This prevents them from hanging around after they finish, but it also means they will close immediately if an error occurs during execution - which makes execution errors difficult to debug because they can no longer be seen when the console window is gone. By enabling install/engine file logging, everything printed to the console window(s) during execution will also be printed to corresponding log files in the tyDiffusion root folder, allowing any errors that occurred during execution to be examined in those logs. The install log file names are: setup.bat.log/setup_ComfyUI.bat.log and the engine log file name is: run_ComfyUI.bat.log.
Root path: this is root folder where tyDiffusion and all of its dependencies (Python, ComfyUI, etc) will be installed. Note: the one-click installer must be run after changes to this path are made.
Model paths (various): these paths define where ComfyUI will search for each model type.
Model paths can be defined with the “$root” symbol, to make them relative to the tyDiffusion root path. Ex: $Root\Models\Diffusion will resolve to C:\ProgramData\tyFlow\tyDiffusion\Models\Diffusion if the tyDiffusion root path is defined as: C:\ProgramData\tyFlow\tyDiffusion.
Some models need to be placed within sub-directories of the base model paths. For example, clip vision models (ex: clip_g.pth from the tyDiffusion HuggingFace repo) for the IP-Adapter module need to be placed in a directory called [External ControlNet models]\preprocessors\clip_vision. The tyDiffusion download manager will set up these sub-directories automatically when downloading models that require them. However, if you don’t use the tyDiffusion download manager to download your models, you will need to ensure they are set up properly yourself. You can see the exact directory structures required for each model type by referencing the following file: [tyDiffusion root]\Engines\ComfyUI\ComfyUI\extra_model_paths.yaml which will be created after tyDiffusion is finished installing and tyDiffusion has run for the first time. Users should not modify that file directly, as changes to it will be automatically overwritten each time tyDiffusion is run. Furthermore, changes made to the external paths in the tyDiffusion UI will only be reflected in that file once tyDiffusion’s engine is restarted.