Source: raw/Google_AI_Can_Now_Turn_Almost_Anything_Into_Content.md (https://www.youtube.com/watch?v=4jZQoxF8YmY) Creator: Grace Leung Platform: YouTube Duration: 16:43 Published: 2026-06-12 Skill level: Beginner to intermediate (no code; Gemini app, Flow and Pomelli web UIs)

Grace Leung builds a content engine for a demo snack brand (“Healthy Crunch”) using only Google’s AI tools, starting from raw brand files: blog posts, a lead-magnet PDF, ad images, product videos and social campaigns. She says she tried the tools first-hand at Google I/O. The stack runs from NotebookLM-in-Gemini (brand memory) through Nano Banana (images), Gems (saved workflows) and the Omni video model, to Google Flow (a video studio with an agent mode) and Pomelli (on-brand social campaigns from your website). The video is sponsored by HubSpot (a free Nano Banana business guide).

Key Takeaways

  • NotebookLM inside Gemini is the brand memory. Notebooks are now reachable from the Gemini app. A “brand knowledge” notebook (hero products, customer insights, product images) acts as an expanding knowledge base that every Gemini task can draw on.
  • Skills work in the Gemini app. She attaches a “blog content brief” skill, Gemini follows its template to brief from the notebook, then Canvas writes the full post with formatting and interlinking. Image mode adds a banner and two inline visuals that stay on-brand because the product photos are in the notebook. Everything exports to Google Docs, visuals included, for review and comments.
  • Nano Banana’s practical marketing moves: pick a pre-built style plus product images for a quick ad shot; edit by circling an area with the sketch tool (“remove this”); ask for three A/B-test image prompts (varied background, mood, style) before generating; and resize a creative to another ratio (square to vertical) while keeping it consistent, done “within 10 seconds” in her demo. She finishes text and slogans in Canva.
  • Gems stop you re-prompting. A Gem is a saved custom Gemini with a name, instructions, a default tool and knowledge files (including notebooks). Her “Studio Ad Shot” Gem uses Nano Banana as its default tool and holds the product packaging dimensions (so outputs are realistic in scale) plus a studio-ad style guide. Make one Gem per repeatable task (studio shots, lifestyle images, campaign copy) and share them with the team.
  • Omni is Google’s any-input video model in Gemini’s video mode. Up to 10 seconds per generation in the Gemini app, strong character consistency, easy scene changes (“Nano Banana, but for video”), and it can take a video as reference for pacing and editing style (10-second clips only, so trim first). Her recipe-reel demo got the recipe steps right.
  • Storyboard before you spend quota. Omni has a strict daily generation limit by plan. She uses a Gem that writes scene-by-scene instructions from Google’s official Omni prompt guide, then generates a storyboard image, and only then renders the video with the storyboard, raw inputs and the correct aspect ratio. She says the output is “much more predictable.”
  • Flow is the studio around the model. It adds a timeline, reusable characters and assets, and controls for video model, ratio, duration and number of variations. A brand character (reference image + prompt + name + voice) can be tagged into any scene. Each generation costs credits (15 in her example) and clips are still 10 seconds, so you extend by adding scenes.
  • Flow’s agent mode orchestrates multiple generations. Give it the product and ask for three ad variations with different camera moves. It can be set to ask approval before acting, and it follows a supplied storyboard with tagged character and product. She says you can’t do this kind of multi-generation orchestration in the Gemini app, and that agent-mode quality is “almost always better.” Mascots work as well as human characters.
  • Pomelli builds on-brand social from your URL. Enter your website to create a “business DNA” (colours, taglines, editable or agent-suggested), check the imported catalog and assets, then either take a pre-built campaign (e.g. World Chocolate Day) or prompt a campaign from selected products, offer, goal and ratio. It writes a creative brief first, then the assets. It also does photoshoots from catalog products (pick four templates), text-only animation, and a simple one-click website for testing a campaign idea. She calls it “a great free tool.”

Try It

  1. Build the brand notebook first. Load hero products, customer insights and product photos into one NotebookLM notebook, then reference it from Gemini for every task.
  2. Make one Gem per repeatable asset. Start with a product ad-shot Gem: Nano Banana as default tool, packaging dimensions and a style guide as knowledge. Share it with the team.
  3. Generate test angles, not just images. Give Nano Banana a product photo and ask for three text-to-image prompts for A/B testing; pick one and generate in the same chat.
  4. Resize, don’t redesign. Feed a finished square creative to Nano Banana and ask for a vertical recomposition that keeps everything consistent.
  5. Storyboard every Omni video. Scene list → storyboard image → render. It protects a small daily quota.
  6. Move multi-clip work to Flow. Create the brand host once as a Flow character, then use agent mode for variation batches.
  7. Try Pomelli for a quick campaign. Point it at your site, correct the business DNA, and generate a campaign brief plus assets. Check you have the rights before publishing its generated website.

Implementation

Tool/Service: Gemini app (NotebookLM notebooks, skills, Canvas, image mode, video mode), Nano Banana, Gemini Gems, the Omni video model, Google Flow, Google Pomelli; Canva for finishing. Setup:

  • Gem: sidebar → Gems → New Gem → name + instructions + default tool + knowledge (files or a notebook).
  • Flow: new project → upload media / create character (reference image, prompt, name, voice) / create scenes → tag character and product in a prompt → choose model, ratio, duration, variation count. Agent mode has its own behaviour settings, including “confirm before acting.”
  • Pomelli: enter website → review business DNA and catalog → pre-built campaign or chat-prompted campaign → brief → assets; Photoshoot and Website sections are separate. Cost: Omni generation limits depend on Gemini subscription plan (not itemised). Flow charges credits per generation (15 credits in her example). She calls Pomelli free. No prices given. Integration notes: Blog drafts and visuals export to Google Docs. Nano Banana and Omni are both reached from mode pickers in the Gemini chat box, each with pre-built styles. The wiki’s Seedance short-film pipeline independently reports Gemini Omni at ~10 seconds, ~2-3 generations a day and no generation API, which matches her limits. That pipeline found Omni weak at 2D-animation character consistency, while she praises its live-action character consistency. The two are different use cases, not a contradiction.

Open Questions

  • Model naming. She calls the video model “Omni” (once captioned “Omi”). The wiki’s other sources name it Gemini Omni. Its relation to Veo, which the HeyGen comparison test ran inside Flow as Veo 3.1, is not explained in the video.
  • Unconfirmed caption terms: the Nano Banana preset she picks is captioned “salon style” (exact name unclear); “photo short” is Pomelli’s photoshoot feature (certain from context).
  • Pomelli availability and pricing (regions, limits, whether “free” persists) are not covered.
  • Skills in Gemini. She attaches a “skill” file in the Gemini app; the video doesn’t say whether that is a native Gemini feature or an uploaded instruction document.
  • “Google has the best video model out there” is her opinion, not a benchmark.