In the world of AI image generation, competition is strong, and apart from solutions from leading names like ChatGPT and Gemini, there are also new and emerging companies looking to grab a slice of the market.
They’re both focused on multimedia, with varying capabilities for generating content like images, videos and other graphics.
The tests
We included five distinct prompts for Leonardo and Midjourney’s AI to handle in order to find out which is the most versatile.
They include a familiar fictional character doing an expected action; another expected action this time being performed by real humans; a slightly more obscure prompt with a standout adjective to steer the AI, an animal doing an unlikely action; and a broader prompt that gives the AI more freedom.
“Create an image of Krampus walking down a street”
Leonardo and Midjourney both created very eerie illustrations of the Central European figure that comes out at Christmas, but Leonardo’s output didn’t offer the most variety. All four images looked remarkably similar, and Krampus was holding similar items each time.
Midjourney’s results were far more appealing, with plenty more variety. The AI also played with lighting and added people to give the images more context. It was substantially quicker to produce its results, too.
Leonardo 28.45s
Midjourney 17.95s
Leonardo 0 – Midjourney 1
“Depict a group of friends hiking up a mountain”
Again, Leonardo’s effort was admirable but repeated four times over. All four options looked remarkably similar and didn’t offer much in terms of variation.
On the flip side, none of Midjourney’s results looked alike. There were photorealistic and painted results drawn to a variety of styles, offering a range of options.
Leonardo also took around 50% longer to produce its results, so the winner is clear.
Leonardo 29.15s
Midjourney 18.43s
Leonardo 0 – Midjourney 1
“Create an abstract illustration of a car”
This time, the result was a lot closer, and the quality of the content generated by both systems was evenly matched. Midjourney had slightly more variation, but both AI image generators produced usable content.
There was also a much smaller time difference between prompt and output.
Leonardo 26.64s
Midjourney 24.07s
Leonardo 1 – Midjourney 1
“Make a lifelike image of a tortoise climbing a tree”
Not the sort of thing you see every day! In fact, the perfect challenge to trip AI up and cause it to make mistakes.
That’s exactly what happened when Leonardo tried to get a tortoise to climb a tree – in three of the four images, the tortoise was hovering mid-air or clinging on by just a leg or two. In isolation, this test suggests that Leonardo might not be able to handle more complex and unique requests, but you’ll need to try your own prompts to see just how un-renderable they are.
Midjourney’s efforts were outstanding and almost realistic, so it certainly gets the point here.
Leonardo 27.08s
Midjourney 21.88s
Leonardo 0 – Midjourney 1
“Show me what it would look like to fly with eagles”
For the first – and last – time, Leonardo’s effort might just be better than Midjourney’s. Even if its consistency in producing four very similar images remains true, we’re able to see a human flying with an eagle.
Midjourney generated three good images, including one of an eagle being ridden by a human, but the third output of a double-bodied, double-headed eagle lets it down, as does the fact that it’s not really showing how what it would be like for a human to fly with an eagle in most pictures.
It’s a shame Leonardo took half a minute to complete the task, but we can overlook that… just.
Leonardo 30.08s
Midjourney 18.63s
Leonardo 1 – Midjourney 0
Midjourney vs. Leonardo: Which is best?
Leonardo 2 – Midjourney 4
Give Leonardo something familiar to produce with minimal need for imagination and it can generally handle the prompt pretty well. We’re not quite sure it needs to produce all four images, though, because they often look very similar. Maybe cutting this down would make the systerm quicker to generate output.
Midjourney proved much better at handling complex requests, and put up a strong fight till the very end.
If we had to pick between Leonardo and Midjourney, we’d take the latter, but if your prompts aren’t going to be overly demanding, Leonardo might still be suitable.
Furthermore, if you need an AI video generator, an art maker and a tool to remove backgrounds from images, Leonardo might be the one for you.
https://cdn.mos.cms.futurecdn.net/L3j6qNqVV8VQCHPBHuqNQb-1200-80.jpg
Source link
