From AI-generated artwork to a more polished vector illustration.
I tested Ideogram Vector by generating an illustration for a lemon ginger tea packaging.
The raw output was a good starting point, but it still needed some work.
So I converted it into an editable SVG, refined the shapes, rearranged and resized the elements, and added a bit of texture to give the illustration more depth.
The editable SVG step is the useful part here—being able to rebalance the shapes and spacing gives you much more control over the packaging. Did you also test how the illustration reads at the final print size?
The jump from Ideogram's raw output to a clean, production SVG is the honest part of any AI-assisted packaging workflow — on lemon ginger tea the biggest remaining call is usually the label hierarchy under a shelf light, so I'd test the after at print size with a dark background...
after for me. shrinking the ginger and letting the half lemon own the right side gives it one clear hero, and the grain takes away that too clean ai look on the leaves
Nice workflow, Karina. Turning the raw Ideogram output into a clean editable SVG is the step that makes it actually usable for packaging, and the tea label looks great.
Such a neat experiment in AI-assisted spatial design! Using 10 photos of a real setup to train a local model like FLUX 3 Klein is next-level prototyping.
Both options look great, but Option A totally wins on the cozy lighting front for me. Check out the post and drop your vote