Amongst all Artificial intelligence Commercially available today Google It is not available to the user only, not in free or paid version. Image creation. This functionality was, in fact, initially there, but was suspended by Google at the end of February Fierce controversies It woke up.
Images created by Gemini”Arbitrary“, that is, without specific references from the user in the line, heavily conditioned by what is actually called ideology”got up“American, it doesInclusion of women and ethnic minorities One of the company's priorities. There were results Contradictory: drawn by Google Gemini Female Popes Obviously Native American raceor players Black Nazis with African facial features.
How Gemini's AI Image Generator Works
Already a few weeks ago Google Reintroduced Image Creation”Business“Gemini available with accounts Google Workspace. However, the functionality is now available For all users.
The Gemini AI graphics engine is now available Figure 3Created at home by Google Deepmind. According to Google's own reports, it is Best model It offers a comparison between them (in those ready for general use) and previous models Best quality images More detail, better lighting, and fewer artifacts and rendering errors (like the famous six-fingered hands).
To use Image 3, ask Gemini (both in the app version and the web version) to draw an object that we describe with text. As with all models of this type, Excellent explanation And The resulting image will be excellent.
All images produced are in resolution 2,048×2,048 pixelsThat is, there is no way to convince Gemini to produce an image that is double, but not square, created by ChatGPT, Dall-E, and Microsoft Copilot (all of which use the same AI graphics model). Mentioning the tax rate also does not help Change shape Final result.
The generation of images, then, It's not very fastIt is definitely slower than OpenAI's Dall-E or other competing models, But at least it's free And currently using Gemini in the free version, there is no maximum limit of images that can be created in a day or a month.
Images return to Gemini, but…
Now Google has refined the algorithm and reintroduced its potential Create images with AI Without getting into the previous problems. Or maybe not: to be honest, it looks like something Google invented today Only half the solution.
Gemini, indeed, at this time It does not allow you to create images of people. So, the problem of conceptual hallucinations doesn't seem to have been resolved, and Google may have re-introduced image creation in order not to pit itself against the competition.
but, No hope Yet entirely his own means, he would have preferred To accomplish the function To avoid recurrence of problems and disputes.
The problem is that Google's ban on creating images of people is at least strangely and concretely enforced. Very aggressive. For example, prompt
Amigurumi turtle walking in the savanna, photographed by a professional photographer, with a blurred background
Google didn't draw any conclusions other than a warning that Gemini could produce images of humans in the future. In fact, to get the turtle plush toy photo, we have to Remove from line the word “I take photos“.
Therefore, Google's artificial intelligence cannot understand when the user is asking to createPicture Off A photographer And when it asks to create onePicture as A photographer.
This is surely a calculated error by Google, necessary to prevent its AI from arriving Handled to ignore blocking In creating images of people.
