A few weeks ago, Gemini released a new image generation model, Nano Banana. The generation and editing the model allows are impressive.
On top of higher quality, the model closely follows text requests, preserves the environment (character, scene, object, etc.), and it's very powerful for editing.
It's able to do:
- Some text-to-image: generate quality images from a prompt
- Image + text-to-image (editing): share an image and use a prompt to add, remove or modify elements, change the style or adjust the colors.
- Multi-image to image (composition and style transfer): several input images to create a new scene or transfer the style from one image to another
- Iterative refinement: editing as a conversation to refine an image over several steps, making small adjustments.
- Text rendering: generating images containing clear, well-positioned text, ideal for logos, diagrams and posters.
You may have already seen fun use cases like this photobooth app:
or with Past Forward (share an image to generate several in styles from different decades)
To use the model, it's simple and freely accessible directly on Gemini :
But you can also use it in Gemini's developer space: Google AI Studio, which is what I recommend. You'll then have more control over generation, which is quite handy.
That's all well and good, but concretely, how can it help you boost your marketing?
Here, the idea is to share a series of examples to give you as many ideas as possible.
Generating static ads
From a simple prompt, you can generate inspiring static images with a logo / slogan. The result is really nice and can be used as a static visual for ads.
But you can also use Google AI Studio to generate an application that produces different scenes from a single image.
Here for example, we take an image of the latest Google Pixel to do it, this one that I found on the web:
With a drag-and-drop into my app, selecting “bus stop ad”, here is the result:
A word about this edition's partner: Brevo 💚
Since we're talking marketing, I'll take the chance to tell you about Brevo : a complete CRM and mailing tool.
Very competitive on the mailing side (pay as you go), it avoids being overcharged as your contact base grows. And you'll no longer get emails like:
CRM management pairs with mailing, and you can set up drag-and-drop automations similar to Zapier.
You can try Brevo for free today or use the code LOUISGRAFFEUIL to get 50% off for 3 months (annual plan).
👉 Head over here if you're interested
Generating avatars / characters / mascots
One of the model's big advantages is creating realistic characters and staying consistent across editing / iteration. It's a major plus and an opportunity to exploit.
This can lead you to create a mascot for your website and stage it on social media, for example. And you can even then create the avatar from a drawing:
The simplest way is to share an image to respect your constraints. If you don't have images, you can generate one that suits you after several iterations.
This can let you create UGC content by sharing an image of your product, for example. I'll come back to it in another edition, it's almost a topic of its own.
The idea is to start from a product and your avatar to stage a scene. Using the VEO 3 model (or another video model), it gives something like this:
The challenge today is audio (in French) and lip synchronization. There are other complexities that I mention here. But today, it's still a lot of post-production work.
Simplifying image editing
No more need for Photoshop, the need is broad and it lets you be more autonomous. It can be removing elements, removing backgrounds, adding filters, etc. A few examples at random:
Multi-reference generation works very well:
Generating videos from your images
Another example here using 3 images to generate a video. We start from a model, a texture (the middle image) and a scene.
We use Nano Banana to generate photos in series that respect the 3 images above. We then get this result:
We use them to generate a video with VEO 3 using the start frame / last frame logic to get 8-second clips, then we assemble everything:
Best practices
Despite the advantages, these AI tools for image creation are still underused. Teams aren't familiar with these solutions and don't know how to make the most of them.
So here are the best practices for better control:
➡️ Be hyper-specific: the more details you give in your prompt, the more control you have over the result.
➡️ Fix character consistency: if a character's features change after several edits, restart the conversation with a detailed description to maintain consistency.
➡️ Give context and intent: a prompt like “Create a logo for a minimalist, high-end skincare brand” will give better results than “Create a logo”
➡️ Iterate and refine: don't expect a perfect result on the first try. Use the model's conversational mode to make small adjustments.
➡️ Use “semantic negative prompts”: instead of saying “no cars”, describe what you want to see, for example “an empty street with no traffic”.
➡️ Control the camera: use photographic terms (like “wide shot”, “close-up”, “low angle”, “85mm portrait lens”) to control the image composition.
