AI workflow for creatives

Workflow with visual references

What a reference says better than text

Late-afternoon light, skin grain, the weave of a fabric, the palette of a brand, a specific face: you can spend ten lines describing it, or show an image.

A visual reference conveys to the engine what text cannot. It is the most natural tool for a creative, because that is already how they brief a photographer or an illustrator.

Every reference has a role

The trouble starts when you give several images without saying which one does what. The engine mixes: it takes the style of one, the subject of another, and returns something that is neither.

A workflow with references assigns a role to each one:

• the subject: this character, this product, this object, that one and no other;
• the style: the material, the rendering, the palette;
• the composition: the framing, the placement of elements, the depth;
• the mood: the light, the time of day, the weather.

Telling the engine which one brings what is what separates a useful reference from one more image in the pile.

The professional case: holding an identity

For a studio or a brand, references serve above all continuity: the same character from one image to the next, the same product across a whole campaign, the same brand guidelines across a series. That is where text alone fails, and where a well-assigned reference holds.

Nova helps organise the references of a project and tell the engine the role of each one. For an identity that has to hold over time, a LoRA is the other way.

Multi-reference engines

Some models accept several input images with distinct roles; others take only one. The page Image editing models covers this.

Want to see what it does on your own visuals? Contact request: a first conversation, free and with no commitment. Every practice is different, and it is by listening to you that we propose what suits you.