AI Image News: The Week of FLUX 3
A quiet week, culminating in one major release. And a tool that disappears.
Figure 01.0 — AI Image News
First, the Terminology
Image Model: the AI that creates an image from text (a prompt). Midjourney, ChatGPT Images, and Google's Nano Banana are well-known examples.
Reference Image: a photo provided as an example. For instance, a face or a product that should appear in the image.
Open Weights: the model itself is downloadable. You can then run it on your own computer.
Arena: a ranking where people blindly compare two images and vote on which is better.
FLUX 3 Image Released
Black Forest Labs released FLUX 3 Image on 1 October. The image component of FLUX 3, which was announced in July.
Creates images directly in 2K and 4K.
You can provide up to ten reference images.
New: with frames, you can determine in advance where something appears in the image and how large it is. This is not yet standard anywhere else.
A free downloadable version (open weights) will follow in a few weeks.
No independent tests yet. We'll know how good it really is in about two weeks.
Reve Stops
Reve, a valued image tool, ceased generation on 27 September.
Reason: an investment from OpenAI. Part of the team moved to OpenAI.
Do you still have images in Reve? You can export them until 31 October. After that, everything will be gone.
Google: Nano Banana 2.1 Coming Soon?
In a new version of Google Flow, the name "Nano Banana 2.1" appears.
Previously, people tested an unknown model under the codename "Spicy Mayo". According to testers, better than the current Nano Banana 2.
Google has confirmed nothing. These are rumours.
OpenAI Tops the List
In the Arena leaderboard (as of 24 September), the new OpenAI models are ranked first and second. Well ahead of the rest.
Microsoft ranks fourth. Google's Nano Banana 2 is only ninth.
At OpenAI's DevDay (29 September), there was no image news.
Small Updates
Ideogram 4.5 was reportedly released on 30 September. Focused on editing in multiple rounds without shifting the image. Not yet officially confirmed.
Midjourney released an interface update. No new model. Edits no longer stretch the image.
Adobe Firefly now supports more image ratios with the OpenAI model.
To Watch
Official announcement of Google's new Nano Banana.
The free downloadable version of FLUX 3 Image.
First independent tests of FLUX 3 Image.
Midjourney plans collaborative image creation, according to the changelog, as early as next week.
