Should product packaging be included in your AI brand guide for video generation?

Yes — if your packaging ever appears on screen, it belongs in your AI brand guide. Include the packaging architecture in detail plus reference photos of every layer from all angles with a scale reference, because AI video models drift on label text, color, and detail unless the packaging is locked in context before generation.

Add a dedicated packaging section to your brand guide before you generate a single shot. Document the packaging architecture in detail — bottle or box structure, label layout, logo placement, on-pack text as an exact string, and finish or material descriptors — alongside your brand DNA, voice, and visual identity. One documented workflow structures the guide PDF with explicit sections for bottle architecture and packaging, and credits that section directly: detailed packaging specs in the guide are what keep the product identical across every generated shot.

Pair the written specs with a packaging reference set. Upload every layer of your packaging from multiple angles, including close-ups — not just one hero image — and include shots of a human hand holding the product and packaging so the model learns true scale. One production calls this a product sheet, and it prevents multiple iterations and wasted credits by solving product drift at the source rather than correcting it shot by shot.

The reason this is non-negotiable: AI video generation causes detail drift and texture shift on complex products, and it worsens as product complexity increases. Without packaging in context, the model invents label text, shifts colors, and redraws layouts between shots — and missing product consistency is one of the three elements that makes an ad read as fake. With packaging locked, one jewelry campaign held 100% product consistency across three complete films (~$2,400 total), and a supplement-brand localization run held 100% product and text consistency across ads for two different markets, validated across 3 ads.

Load the guide once, not per prompt. invideo is an agentic video creation tool where the invideo agent holds a persistent context tab — the project's memory — so a brand guide uploaded at the start applies to every image and video generated afterward without re-prompting. Initial brand context setup takes about 15–20 minutes, and documented productions found that after that one-time load they never re-explained packaging rules on any generation. Where image models come into play, a two-stage pass — GPT-Image-2 for the aesthetic base, then Nano Banana to lock exact product detail into the frame — leans directly on those packaging references, and the invideo agent routes between the models for you.

Packaging inclusion pays off most in three situations: the pack is visible in your video shots; you run multiple SKUs or localize winning ads across markets and need the product identical in every variant (one localized-ad run generated 27 images per ad while holding the pack and on-screen text consistent); or the product itself is intricate, where drift is worst. If your packaging genuinely never appears on camera — a pure service brand, for instance — you can leave it out of the guide, but for any physical product the reference set costs 15 minutes and removes an entire category of failed generations.

Watch some of these to see what works for you:

Put your bottle architecture and packaging in the guide in detail. That's what keeps your product identical across every single shot.

— invideo's creative team