When developing AI imagery, I want to keep the process as well as the selected image. A different angle or a quieter background is easier to discuss when the original settings are still available.
ComfyUI connects models, reference images, prompts and outputs as nodes. Workflows can be saved and reopened; the official introduction includes a complete example.
When I would consider it
It makes sense when a direction needs repeated adjustments or several images share the same setup. I would keep what is working and compare one substantial change at a time.
For product imagery, I would inspect the silhouette, ports, labels and proportions separately from the mood. A complete node graph does not guarantee an accurate product.
Start with one image
Run an official example before adding a long chain of custom nodes. Keep the starting workflow and compare outputs side by side. A useful handover also identifies the models, nodes and reference assets it needs.
The time it needs
Local use involves hardware and dependency setup. Cloud services involve their own charges and handling of assets. Open-source software does not make every model or generation service free.
For a single quick mood image, setup may take longer than the creative experiment. I would first decide whether this process needs to be repeated.