3 workflows for 'Nano Banana' (Google's new AI image model)

Hello everybody, welcome back to Moodboard 🌴

“Nano Banana” (Google’s new, updated AI image generation model) has gotten a lot of buzz since the launch earlier this week.

I’ve been playing around with it for the last couple days, and suffice it to say that I’m very impressed.

I don’t want to get too ahead of myself, but the image editing capability makes it feel like the “AI Photoshop” solution that a lot of us have been waiting for.

In today’s edition, I’ve outlined some of the more powerful initial workflows/use cases that I’ve been experimenting with.

Ready? Let's dive in 👇

What is “Nano Banana?”

Anyone that has spent time using AI image-generation tools knows that making small, specific, accurate edits to images via prompting has been (up to this point) near-impossible.

The AI has always changed something you don’t want changed, altered the image in a way you didn’t ask for, or simply not made the edit you requested.

Well, I’m happy to report, it seems that Nano Banana largely solves these issues 🙌

Gemini users (even on the free tier!) now have access to an AI image editor that generates professional-quality results, with features like:

• Language-based editing: Tell it what you want changed in plain English, and it will carry out specific, accurate edits to your image

• Identity preservation: Characters and objects stay consistent across multiple edits and angles 🤯

• Speed: Most generations complete in ~2-3 seconds.

Having experimented with it quite a bit over the last couple of days, I can say that it really does feel like a huge unlock.

To demonstrate just how powerful this new model is, I’ve assembled a few use-cases for marketers, operators, and business owners that want to start getting value from it right away.

Workflow #1: The Product Showcase

One of the most immediately useful features of this new model is that it allows you to accurately swap products into any scene or existing image. 

This obviously unlocks massive opportunity for marketers and content creators who want to showcase physical products in different contexts.

It’s as simple as:

1. Uploading a product image (and any additional images of backgrounds/characters/etc that you want to be included)

2. Describing the scene/context you want the product to be showcased in

3. Watch Nano Banana work its magic

As a compulsive drinker of sparkling water (I blame years of SF office culture), I decided to test this kind of workflow out with a can of Spindrift.

I grabbed a product photo of the can, pasted a Moodboard-esque shot of a man holding a soda can that I generated in Midjourney, and let Gemini work its magic.

Here are the two images:

The only prompt that I included was “Swap the soda out for the Spindrift can.” 

Here is the Nano Banana output:

Notice how It didn't just copy/paste the photograph of the product in. It adjusted shadows, art style, and lighting to match the scene perfectly.

Think about all the ways that creators & marketers can use this functionality. The possibilities are too many to list out!

Workflow #2: The Aesthetic Collage

I saw someone on X demonstrate this, and I had to try it myself (and share it with you all).

Nano Banana is so good at accurately replacing/inserting individual elements that you can upload multiple unrelated images, and Nano Banana intelligently combines them into a cohesive almost ‘collage-style’ composition.

To try this out, I gathered 8 images that kinda fit Moodboard's beach-nostalgia/vibey aesthetic: 

Some Pit Viper sunglasses, a walkman, a 90’s-era Volvo 940 (my first car), some Japanese city pop art, and a picture of Timothy Olyphant (I’ve been binging Deadwood lately).

I uploaded them all into Higgsfield chat (using the Nano Banana model) and prompted it to combine all the elements.

This is the output it gave me 👇

Pretty impressive!

Just look at the accuracy with which it replicated each individual element and the resulting coherence of the output. This is something we haven’t been able to accomplish with AI image-gen up to this point. 

This allows for much more detailed, nuanced design and styling of the AI-generated images we create. 

We can now truly manage (and eliminate) the many little frustrations that kept AI images from reaching that polished/finished state through a prompting workflow alone.

Again, the possibilities are seemingly endless.

Workflow #3: AI Video (Using Start + End Frames)

This is, in my opinion, where Nano Banana unlocks some real magic, especially for folks doing a lot of work with AI video tools.

All of the most popular AI video generators (Runway, Veo, Higgsfield, Midjourney, etc.) allow you to have increased control over the output by setting “start” and “end” frames.

The challenge, however, has always been creating end frames that match the details within your start frame while showing progression in the narrative.

For example if a character’s face looks different in the end frame relative to the start frame, the video output will need to compensate for that and it will morph the character’s face over the course of the video. 

This, of course, looks strange/surreal and gives a distinct ‘AI look’ to the videos we generate.

Nano Banana solves this problem with massively improved coherence/consistency across image generations. 

Let me show you what I mean.

I created this image (featuring a suave character walking along the beach) in Midjourney.

Now, say that I want to generate a video where he sits down on the brick walkway in the photo. 

To accomplish this, I want to make an ‘end frame’ that shows him sitting down over there.

Historically, I would be worried that a request to recreate/edit the image would change a bunch of other aspects of the shot. It might remove his hat for example, or change how many trees are in the background.

With Nano Banana, we really don’t have to worry about that. I asked for a copy of the image with him sitting, and this is what it gave me:

Notice the continuity! Everything else in the image has remained constant, except for the one edit that I requested (it even got the angles of the shadows right?!).

Because of the execution here, I can create a video in Midjourney that looks a lot more polished and consistent. Here’s what I got 👇

Give It A Try

All in all, this feels like a huge leap forward for the usefulness of AI image generation as it pertains to marketing/content creation.

It feels like we’re that much closer to getting AI image assets that are 100% “done” through prompting alone.

As you might imagine, there are infinitely more use cases for this update across marketing and content creation. I’ll continue to share the results of my own experiments here in Moodboard.

Play around with it yourself, and let me know what you end up creating!

Until next week,

Ryan