ComfyUI Workflow Examples for Browser Editing
These are structural starter workflows, not claimed benchmark results or tested renders on specific hardware. They use core ComfyUI node names and placeholder filenames. Edit them in Workflow Studio or load them in your own compatible ComfyUI installation.
Text-to-image starter
CheckpointLoaderSimple supplies the model, CLIP and VAE. Two CLIPTextEncode nodes provide positive and negative conditioning. EmptyLatentImage sets dimensions and batch size; KSampler samples, VAEDecode decodes, and SaveImage writes the result.
Image-to-image starter
LoadImage and VAEEncode supply image latents to the sampler. The denoise value begins at 0.65 and changes how much of the source image is preserved. Set the image filename to a file available in your ComfyUI environment.
Before running either workflow
- Choose a compatible checkpoint you actually have installed.
- Replace the positive and negative prompt text.
- Confirm dimensions and batch size fit your available memory.
- For image-to-image, make the source image available in ComfyUI.
- Use the structural checker, then test the workflow in ComfyUI.
The exported format is legacy UI JSON v0.4. These examples do not cover every model family, specialized custom widget or newer subgraph format. See the official ComfyUI workflow format.