HKF Group: A Controlled GenAI Production Workflow

Corporate image work often begins with fixed architecture, approved source objects and a composition that cannot drift. Unconstrained generation can produce attractive images quickly, but it is unreliable when placement, brand detail and repeatability matter.

I designed a controlled ComfyUI and Flux workflow that locks the composition before generation, uses masked source assets, and keeps manual review at every decision point. The workflow was tested for selected tasks; it is not presented as a company-wide deployment.

The composition is built first. Environment photography and approved reference objects are placed with spatial masks so scale, position and the main perspective are established before model processing.

This reduces random variation and gives the designer a more predictable base for subsequent lighting, reflection and material work.

The test combines an interior corridor, placement guides and a mechanical-clock reference. Masking establishes where the new object belongs; image-to-image processing resolves the surrounding light and material relationships.

The reviewed output demonstrates the value of the sequence: the model assists integration, while composition, scale, brand suitability and final quality remain designer-led.

The result is a repeatable production path with clear inputs, review points and handoff stages—not a one-click image generator.

The project reframes GenAI as one controlled stage within a larger design workflow. Its value comes from locked inputs, deliberate review and accountable handoff.

For selected corporate visual tasks, that structure can accelerate exploration without giving up composition control, source-asset discipline or human quality decisions.