What should a lookbook PDF include for accurate AI fabric generation?

A lookbook PDF for accurate AI fabric generation should carry, per garment: a fabric close-up of the weave, front/side/back shots, an on-model shot, and tagged metadata (look number, fabric type and weight, composition, colorway, silhouette). Multiple angles and a fabric macro are the non-negotiables — variety of images is what lets the agent render the cloth correctly across shots.

Build the PDF garment by garment, and for each garment include these inputs:

1. Fabric close-up (macro on the weave). A tight shot showing the weave structure, sheen, and surface — this is what teaches the agent how light hits the cloth. Without it, fabrics drift toward generic plastic-looking AI rendering.

2. Front, side, and back product shots. Three orthogonal angles on the garment isolated (no model) so the agent has a clean silhouette read from every side. This is the same input pattern that holds for product reference in general: close-up, front, side, back.

3. On-model shot. At least one full-look shot of the garment worn, so the agent knows how the fabric falls on a body — drape, break, where it gathers, how it moves at the shoulder and hem.

4. A garment-fall shot where movement matters. For anything where drape sells the fabric (linen, silk, knits, heavy wool), add a shot showing the garment in motion or hanging — this communicates how the fabric actually feels to a viewer.

5. Tagged metadata per look. Every look gets a structured header the agent can read: look number, garment name, fabric type and weight (e.g. "linen, 180gsm"), fiber composition, colorway, silhouette descriptor. Missing fabric/silhouette tags is the most common cause of fabric drift between shots — the agent has the pictures but no label to anchor them to.

6. Multiple angles per garment, not one hero each. Variety of images is the critical input. One image per look is not enough; the agent needs enough visual coverage that it can reconstruct the garment from any camera angle the shot list calls for.

One orientation note: invideo is an agentic video creation tool, and the lookbook PDF is uploaded once into the invideo agent's context tab alongside your brand brief and shooting script — it then governs every generation in the project without you re-prompting wardrobe per shot. If you don't have a lookbook PDF built yet, you can jam with the invideo agent to generate one for your collection before production.

What the lookbook PDF does NOT need to carry: the verbal Texture Language descriptions (how each fabric feels in words — reflectivity, temperature, organicness) live in the agent brief, and per-shot fabric-movement direction lives in the shooting script. Keep the lookbook focused on visual reference and metadata; let the other two documents carry the verbal direction.

Validate before you commit the run. Once the PDF is loaded, generate one garment across a close, mid, and wide before batch-generating the campaign. If the weave, color, and drape hold across all three distances, the lookbook is doing its job; if any one drifts, the fix is usually a missing angle or a missing metadata tag for that look.

Watch some of these to see what works for you:

It's important to give the agent a close-up of the fabric, a front shot, a side shot, a back shot, and the fabric worn on a person.

— invideo's creative team