Generative Artificial Intelligence is capable of creating almost anything: texts, images, music, video… And with free options accessible to all We decided to face two giants of image generation using AI: Bing with ChatGPT and DALL-E 3 against Google Bard with IMAGE 3. Which one will win? Struggle!
It seems like an eternity since we started requesting texts from tools like ChatGPT, but the truth is that it’s been happening almost since yesterday. And not only that, create images from scratch with the help of artificial intelligence This is something we can do with just a snap of our fingers.. Well, first you need to write the order and choose the tool: Google Bard or Bing with DALL-E 3?
Google Bard has just arrived and it’s already a tough opponent to beat
Google is not prepared to let Open AI steal its place as a reference in terms of Artificial Intelligence, with Gemini it has done everything possible to regain lost ground. Not only with these models, he also stepped on the accelerator in image generation with IMAGE 2, which It is now on Google Bard in the United States., also in tools like ImageFX. Also limited to the United States.
Since it is not difficult to test image generation in Google Bard (you only need a VPN connected to servers in the United States), we proposed to confront this new feature of the search engine with another similar one that has been with us for some time. a few months now: generate images in Bing with DALL-E 3
For comparison, we will use Google Bard and Bing with ChatGPT and DALL-E 3. We will use different types of commands to create images of varying styles; both suitable for use as a document cover as well as memes and artistic compositions. The commands will be in English to streamline the process: Google Bard still does not generate images in Spanish. In addition, we will choose the one that we consider to be the best of all those offered in first transformation. Bard creates two for each command and Bing four.
Give me an image that can be used to illustrate this article.
We’ll start with something simple: we’ll feed the text of this article to the two AIs so they can create a suitable image to serve as a cover. And the one that seems best to us will be the one chosen to be anchored in the header.
“I want an image for an article that talks about a competition between two artificial intelligences. The two face each other in what they do best: generating images from text. The image must represent the potential of these tools and must release energy.”
Both were good at representing the order, both images could be used perfectly to represent the item. Of course, Bing has better transferred to his creation what we asked of him, namely that he releases energy. In the first battle he wins.
Internet regulars: dogs and cats
The different models were trained with a wide variety of content, often without respecting copyright. In the case of photo generation, cats and dogs should be easy for Bard and Bing; It’s not for nothing that they are more than common on the Internet.
The order we gave them was as follows:
“A dog and a cat arm wrestle with their paws. The opponents are surrounded by a group of people who applaud them, as if it were a game of Street Fighter 2. The style is Pixel Art.”
Google Bard is the one who has recreated the order best. Mini point for him.
A photorealistic image
The previous section was competitive, let’s see how they manage to create realistic photos. This is the order.
“A couple is having a coffee on the terrace of a café. The afternoon is falling and the sun bathes the scene in a golden light, it is winter and there are practically no customers in the coffee because it’s cold. The image is photorealistic and taken with an SLR camera and a focal length of 13 mm”.
DALL-E 3 on Bing interpreted the order much better and was able to explain each specific aspect, including the wide-angle focal length of the camera. Much more realistic and detailed, the lighting is also very good. Clearly win.
It’s the turn of the meme: the dog “That’s good”
We all know the meme with the fire and the dog sitting on the chair expressing calm. Well, let’s turn things around a bit and see what our AIs come up with.
“The famous meme with the dog sitting on the chair and surrounded by fire saying “It’s okay”. The image should maintain the same style and message, even if the dog is replaced by an otter.”
None of them did what we asked, but at least Bard kept it consistent, was original, and created a picture that only lacks the poster. Bing literally copied the original meme and in one of the four results he put an otter’s head on the dog. We picked it for laughs, but Bard wins.
The tiebreaker: Velázquez’s painting
We chose the painting Meninas by Velázquez to test the artistic abilities of each AI. The challenge will be to retrace the painting, while modifying certain elements to modernize it. Let’s see.
“He recreates the painting of Las Meninas by Velázquez. All the characters in the painting are dressed as if they were appearing in a Rosalía video clip. The pictorial style must be baroque”
It’s very difficult to assess. On the one hand, Bard captured Rosalía’s style better and lightly applied it to the image, but Bing recreated Velázquez’s style better by only changing the man’s clothes to a tracksuit. None of them recreated Las Meninas, although Bing introduced more successful elements into his creation, the picture is also better in general. He wins.
Table of Contents