Back in November, I tested the photo gearing features Within Google’s Gemini, run by the Imagen 3 model. While I liked it, I ran into the limitations pretty quickly. Google recently rolled out the successor – Imagen 4 – and I have put it through the pace for the last couple of weeks.
I think the new version is definitely an improvement, like some of the problems I had with Picture 3 Fortunately, is now gone. But there are still some frustrations, which means the new version is not as good as I want.
How often do you make pictures with AI?
11 votes
So what has improved?
The quality of the images produced has generally improved, although the improvement is not massive. Imagen 3 was already generally good at creating pictures of people, animals and nature, but the new version consistently produces sharper, more detailed images.
When it comes to generating images of people-who is only possible with Gemini Advanced, I had persistent problems with Imagen 3 where it would create cartoon-looking images, even when I did not ask for the specific style. Asking it to change the image to something more realistic was often a losing battle. I haven’t experienced any of it with Imagen.
One of my greatest frustrations with the older model was the limited control over side conditions. I often felt stuck with 1: 1 square images, which limited the use of use. I couldn’t use them for online publications, and printing them for a standard photo frame was out of date.
While Imagen 4 is still standard for a 1: 1 ratio, I can simply ask it to use another, as 16: 9, 9:16 or 4: 3. This is the function I have been waiting for, as it makes the images created far more versatile and usable.
Imagen 4 also works much smoother. Although I have not found it to be noticeably faster – although a faster model is reportedly in the works – there are far fewer errors. With the previous version, Gemini would sometimes show an error message and said it could not produce an image for an unknown reason. I have received none of them with Imagen 4. It just works.
Still looking a little to retouched
While Imagen 4 produces better images, is more reliable and allows for different aspect conditions, some of the problems I have experienced when I test the predecessor are still present.
My main problem is that the images are often not as realistic as I want, especially when I make close -ups of people and animals. Pictures tend to get out quite saturated, and many have a prominent bokeh effect that professionally blows the background. They all look like they were taken by a photographer with 15 years of experience instead of me, just pointed a camera at my cat and pressed the shutter.
Sure, they look nice, but a “random mode” would be a fantastic addition – something more realistic, where the lighting is not perfect and the topic does not posing as a model. I got Gemini to make a picture more realistic by removing the bokeh effect and generally making it less perfect. AI tried, but after asking it three or four times in the same picture, it seemed to reach the limit and said it couldn’t do any better. Each new image it produced was a little more random, but it was still quite polished, and clearly suggested that it was AI-generated.
You can see that in the pictures above, go from left to right. The first includes a strong bokeh effect, and the man has very clear skin, while the other two move on to the man who sees the elderly and the elderly, as well as more tired. He even started bald a little in the last picture. That’s not what I really meant when I ask Gemini to make the image more realistic, even though it comes out more randomly.
Imagen 4 does a much better job with random images such as landscape and urban rins. These images, taken at a long distance, do not include so many close -ups, so they look more real. Still, it can be a hit or miss. A picture of the Sydney Opera House looks good, although the saturation is supported quite a bit-grass is extra green, and the water is a perfect picture. But when I asked for a picture of the Grand Canyon, it came out completely artificial and wouldn’t fool anyone into thinking it was a real picture. It performed better after a few attempts.
Editing is better but not quite there
One of my grips with the previous version was the clumsy editing. When asked to change something smaller – like the color of a hat – would AI do it, but it would also generate a whole new, completely different image. The ideal scenario would be to create an image and then be allowed to edit each detail accurately, such as changing a clothing, adding a specific item or changing the weather while leaving everything else as it is.
Imagen 4 is better in this regard, but not so much. When I got it to change the color of a jacket for blue, it created a new image. However, by specifically asking it to keep all other details the same, it managed to maintain a lot of nature and topic from the original. That’s what happened in the above examples. The woman in the third picture was the same, and she seemed to be in a similar room, but her pose and camera angle were different, which made it more to shoot again than an editing.
Here’s another example of a cat eating a popsicle. I got Gemini to change the color of Popsicle, and it did, and it kept a lot of the details. The cat is the same and it is most of the background. But the cat’s ears are now sticking out, and the hat is a little different. Still, a good try.
In spite of the deficiencies, Imagen 4 is a great tool
Even with the problems and a long wish list with lack of functionality, Imagen 4 is still among the best AI image generators available. Most of the problems I have mentioned are also present in another AI image generation software, so it is not as if Gemini is behind the competition. It seems that there are significant technical obstacles that need to be overcome before this type of tool can reach the next level of precision and realism.
Other restrictions are still in place, such as the inability to create images of famous people or generate content that breaks Google’s security guidelines. Whether it is a good or a bad thing is a matter of meaning. For users seeking fewer restrictions, there are options that Purse.
Have you tried out the latest photo generation in Gemini? Let me know your thoughts in the comments.