Google AI Can Now Turn Almost Anything Into Content

Grace Leung · 20:07 · one demo brand (fictional snack “Healthy Crunch”), zero raw files → full content engine · watch

Marketer’s walkthrough of the connected Google AI stack, run end-to-end on a single demo brand: blog article + banner, lead-magnet PDF, ads, product videos, social campaign. (Middle segment sponsored by HubSpot’s Nano Banana guide — flagged.)

1. NotebookLM is now inside the Gemini app

• Brand knowledge notebook = project space that remembers context: hero products, customer insights, product images.
• Attach a content-brief template → Gemini drafts the brief from notebook sources, then Canvas mode writes the full post with formatting + interlinks.
• Image mode generates an on-brand feature banner + supporting visuals using the product images stored in the notebook.
• One-click export to Google Docs (with visuals) for team comments — still review titles/copy before publishing.
• Same notebook also powers a lead-magnet PDF draft (parents’ lunchbox planner) refined in Canva, and internal docs like a retail line sheet (~10 minutes).

2. Nano Banana (image generation) for marketers

• Pre-built styles in image mode → feed product shots, get campaign-grade ad visuals even from thin prompts.
• Iteration workflow: click the image → sketch-tool circle the region → remove/change it, keep looping; finish polish in Canva (slogan overlays etc.).
• A/B ideation: give the product image, ask for three text-to-image prompts varying background/mood/style, then generate the winners in the same chat.
• Killer feature: creative resizing — recompose a square creative into vertical/other ratios while keeping everything consistent, ready in ~10 seconds.

3. Jams — custom Geminis per repeatable task

• Sidebar → Jams → New: name + instructions define behavior; pick default tools and knowledge sources (product packaging dimensions make studio shots more realistic; style guides or entire notebooks can be attached).
• Example jams: “Studio Ad Shot” (any product in → consistent studio visual) and lifestyle-image generation; shareable across the team so nobody rewrites prompts.

4. Video mode (Veo-class model)

Auto-captions garble the model name (“me/Umi/Omi”) — this is Google’s video model in the Gemini app:

• Any-input→video-output emphasis: strong physics/world reasoning, character consistency across scenes (same presenter throughout).
• Up to 10-second generations in-app; scene changes via prompt like Nano Banana for video.
• Strict daily quota → storyboard-first workflow: a dedicated jam generates scene-by-scene instructions using Google’s official prompt guide + a storyboard image, so you fix details before burning generations; then feed storyboard visuals + raw input back for the final render.
• Also accepts reference videos (trim to ≤10s first): borrowed pacing/editing of an IG reel recipe clip, applied to product + real recipe instructions — output got the recipe right.

5. Google Flow — the studio wrapper

• Timeline editor, reusable characters/assets, scene chaining where the Gemini app stops.
• Brand characters: create once from a reference image (+name, picked voice), tag them + your product in any scene — strong consistency incl. the product itself.
• Control over model/ratio/duration/variations; each generation costs credits (~15 for his example). Standard outputs are 10s — extend by adding scenes.
• Agent mode: orchestrates multi-step video tasks — e.g. “three variations of this ad with different camera moves,” asking approval before acting; excellent at following storyboards when you hand it one plus tagged character/product.

6. Pomelli (captions say “Pomona”) — automated branded social

• Reads your website → builds “Business DNA”: colors, taglines, auto-collected product assets; agent suggests refinements.
• Two campaign paths: pre-built occasion campaigns (World Chocolate Day-style) with editable headers/CTAs/text-animation, or checkbox flow — pick products, describe offer, choose ratio → generated creative brief → full asset set.
• Photo-shot templates from catalog items (pick layouts from a library of four), one-click download-all.
• Bonus: quick throwaway website generated from the brand DNA — not custom-site grade but fine for testing a campaign idea; verify rights before publishing anything.

Takeaway shape

The pattern across all six: context once (notebook/jam/DNA), generate many — consistency comes from stored brand state, iteration from Nano-Banana-style editing, predictability from storyboards before you spend quota.