Follow ZDNET: Add us as a preferred source on Google.
ZDNET’s key takeaways
- Black Forest Labs launched its Flux 2 image generator.
- Benchmarks show that it performs competitively.
- Everyone can try it for free.
If you feel like AI text-to-image generators have become so capable that it’s increasingly difficult to distinguish fiction from reality, that’s only going to get more challenging.
Also: Is that an AI video? 6 telltale signs it’s a fake
On Tuesday, Black Forest Labs, a German start-up best known for its suite of Flux 1 models, released its next-generation model, Flux.2 — right on the heels of Google’s most capable model yet, Nano Banana Pro. Flux.2 features upgrades on nearly every front, resulting in better-quality images that more closely adhere to your prompts.
What Flux.2 offers
The model family includes Flux 2 Pro and Flux 2 Flex, with the former offering the highest quality and the latter providing more flexibility between speed and quality.
Some specific improvements of the model include support for up to 10 reference images, meaning you can incorporate a lot more elements from different pictures in your final product; improved photorealism and detail; more accurate text rendering, a task image generating models frequently struggle with; better prompt following; and a better understanding of real-world knowledge, according to Black Forest Labs.
Also: This new most popular AI image and video generator has enterprise users flocking to it
If these updates all sound familiar, it’s because Google’s Nano Banana Pro flaunted the same upgrades when it was released last week. In fact, Black Forest Labs published benchmarks that compared its performance to that of leading models, including Google’s. While Flux 2 Pro had an ELO score (a rating used for head-to-head model performance) slightly below that of Nano Banana 2 (another term for Pro), the cost of Flux’s model is lower compared to the much higher cost of the Nano Banana model, as seen below.
The cost-per-quality ratio is particularly important for developers who want to incorporate a high-quality image-generating model into their own projects at a lower cost.
My experience (and how to try it yourself)
All users can try the models for free on the BFL Playground, with a limit of 50 renderings per user per day.
For a first stab at how the model could perform, I entered the prompt: Vibrant realistic living turkey. I got the result you see at the top of the article, which did a decent job interpreting the prompt with vibrant colors, details, and a clear representation of a turkey.
I then compared Flux.2 to Nano Banana Pro, which you can also access for free on the Gemini app, albeit with limitations, and was impressed by the comparison – you can see both results side-by-side below.
Also: I test AI tools for a living. Here are 3 image generators I actually use and how
The prompt I used for both was: a Thanksgiving dinner table spread with vibrant colors, rich details, and warm/cozy lighting. Include a turkey at the center, cranberry sauce, and Brussels sprouts. I also want a menu that lists all of these elements pictured.
The prompt was challenging because it didn’t explicitly state what I wanted included, but both generators did a decent job of assuming what I wanted printed on the menu and included all the elements I mentioned. Notably, the Flux 2 Pro model was quicker, though the text was much better on Google’s version.
Artificial Intelligence
<!–>
–>
