👨‍🚀 Edit images perfectly with simple text prompts

In partnership with

Hi AInauts,

Welcome to the latest issue of your favorite newsletter! .

Here's what we have in store for you:

• 🖼️ Edit images perfectly with simple text prompts

• 🔥 New standards in AI video tools

• 🧮 Copilot Formula: More AI for Excel

Let's go!

Find out why 1M+ professionals read Superhuman AI daily.

In 2 years you will be working for AI

Or an AI will be working for you

Here's how you can future-proof yourself:

1. Join the Superhuman AI newsletter – read by 1M+ people at top companies

2. Master AI tools, tutorials, and news in just 3 minutes a day

3. Become 10X more productive using AI

Join 1,000,000+ pros at companies like Google, Meta, and Amazon that are using AI to get ahead.

Sign up and start learning AI

🖼️ Edit images perfectly with simple text prompts

Okay, creating images with AI? No problem. That works. Anyone can do it. Got it.

But editing existing images with simple text? This is where it gets a little more challenging. It doesn't always work.

A new "incognito" image editing model is currently going viral on LMArena. It is called Nano-Banana and is rumored to have been developed by Google. In addition, Alibaba recently released a super exciting open-source model called Qwen Image Edit.

This is reason enough for us to revisit the topic of image editing with AI and take a look at some cool possibilities.

Nano-Banana: A still-secret model from Google?

Let's start with the model that is currently being hotly debated on LMArena and X. It supposedly comes from Google and is great at combining and modifying images.

Want some examples? Take a look at this thread on X:

Want to try it out for yourself?

The whole thing is still in test mode and has not yet been officially released, so it currently only works in LMArena.

So go directly to LMArena, click on Image, upload two images, enter some text, and start the battle.

But one important note: This does not work in a side-by-side comparison. You may have to do a few battles before you get results from nano-banana.

Alibaba with new image editing model

For all open source fans, there is also good news: Alibaba has introduced the new Qwen Image Edit!

The results are quite impressive. Whether you want to maintain a character as shown above or edit existing images as shown below.

During our tests, we noticed that you have to be very precise with your prompts, otherwise the results are rather mediocre.

The exciting thing about Qwen is that you can really edit step-by-step images.

And thanks to open source, it's also extremely inexpensive, so you really don't have to worry about the cost if you spend a little longer experimenting to get the perfect image.

How to try out Qwen & edit images

There are many ways to try out Qwen. But we like Replicate best! Here, you can get images edited for as little as 3 cents per image.

Using the Replicate Playground, you can also compare Qwen with the German rock stars from Black Forest Labs. The Flux Context models have nothing to hide.

You can also use this to edit images using prompts with the FLUX model (see our article here).

Below is a comparison with our Nauti character and a simple prompt.

We’ve achieved the best results with Flux Kontext Pro. Not at all with Flux Kontext Max. Which is great, because Max is twice as expensive as Pro!

Qwen failed the test above quite badly; it needs much more detailed prompts like Flux. It's best to try them all out on your own use case!

By the way, the new nano-banana seems to be even better than the current models. Until then, our favorite remains Flux-Kontext-Pro.

Conclusion: Why is this important? What are the use cases?

We often want to edit or combine existing images. But we don't necessarily always want to generate new ones.

Until now, this has been quite a laborious process. New models are making it easier and better, and even complex multistep editing processes are now easily achievable. Even without AI, changing the entire image.

For us, the best use cases are:

• Combine product shots from two images

• Show brand assets in different situations without losing CI

• Modify text in graphics

Note: Unfortunately, all models are still extremely prone to errors when dealing with German texts. Therefore, brevity is often the key! 😉 

🔥 Higgsfield sets a new standard in AI video tools

Okay, after all this talk about images, we have to mention our favorite AI video tool. We are, of course, talking about Higgsfield AI.

The new company founded by Snapchat's former AI chief is doing exactly what you would expect from such a constellation.

Great AI features, really impressive.

We have written about Higgsfield several times before. Here and here, for example.

A new feature has just been introduced that delivers absolutely impressive results: Draw-to-Video.

The idea behind this is to simply write instructions, draw, upload images, and create dynamic videos based on a blank canvas.

Take a look at the examples in the thread above: whether multi-stage blockbuster visuals, creative experiments, or our favorite use case, product placement videos, anything goes!

Here's how it works:

Unfortunately, this feature is currently only available with a paid plan. But as always, there is a monthly subscription that is not too expensive and can be canceled at any time.

• Create an account here

• Draw/scribble instructions directly into the frame

• Select your model (Minimax has unlimited Creations this week)

• Generate and marvel

Conclusion: Really cool, have fun experimenting!

We don't know exactly how Higgsfield does it. But in our eyes, these guys and girls are constantly setting new standards in video production.

We are excited to see what else is in store. Give it a try!

🧮 Copilot Formula: More AI for Excel

Finally, a brief topic, more of a note.

We don't write much about Microsoft, Office, and Copilot. That's because we don't use them much and haven't really been impressed by them so far.

But many are (still) trapped in the Microsoft Copilot world through their companies.

So here's a quick positive update: You can now control Copilot directly in Excel using a formula, similar to the =ai function in Google Sheets.

The whole thing is still in beta, so it will be a bit shaky at times.

You can now access the beta feature starting with version 2509 (build 19212.20000) for Windows and version 16.101 (build 25081334) for Mac. With the appropriate license for Microsoft 365 Copilot.

The syntax is simple, just like Google:

=COPILOT("Prompt", Range)

Definitely worth a try if you work with tables a lot!

We made it! But no need to be sad. The AInauts will be back soon, with new stuff for you.

Reto & Fabian from the AInauts

P.S.: Follow us on social media - it motivates us to keep going 😁!
X, LinkedIn, Facebook, Insta, YouTube, TikTok

Your feedback is essential for us. We read EVERY comment and feedback, just respond to this email. Tell us what was (not) good and what is interesting for YOU.