Best AI Image Generator Tools (Tested & Compared)
- Sean Barber
- Apr 18
- 2 min read
I often use AI to create images, so I decided to test a few different tools to see which one gave the best results for three different prompts. The tools I tried were ChatGPT, Sora, Mage, Craiyon, DeepAI, and Leonardo.ai.
All of these tools are free to use, except for Sora, which is only available as part of the paid ChatGPT version (ChatGPT Plus).
Prompt 1: A realistic photo of a person walking a dachshund (sausage dog) on a sunny city street, wearing casual clothes, clean pavement, leafy trees, cars parked along the road, early morning light, natural shadows, photorealistic, DSLR quality.
Prompt 2: A realistic futuristic vision of London in 2035, with advanced public transport, green buildings, smart traffic lights, a mix of historical and modern architecture, the Shard in the distance, eco-friendly tech, digital signage, clean air, pedestrians and cyclists, photorealistic, ultra high definition.
Prompt 3: A high-resolution, realistic photograph of the Hollywood sign in Los Angeles, sunny afternoon, clear blue sky, seen from a nearby hillside with trees and dry grass, shadows cast on the letters, cinematic angle, DSLR camera quality.
See how they performed below >
Prompt 1 Comparison
Sora

Deep AI

Canva

ChatGPT

Mage

Craiyon

Leonardo.ai

Prompt 2 Comparison
Sora

Deep AI

Canva

ChatGPT

Mage

Craiyon

Leonardo.ai

Prompt 3 Comparison
Sora

Deep AI

Canva

ChatGPT

Mage

Craiyon

Leonardo.ai

Firstly to note there are other options that are either free or relativly cheap, however, for now I selected these, happy to hear about others so please do share.
Verdict:
To be honest, I expected strong results from ChatGPT and Sora given all the recent updates. Even when Sora generates just a single image (like a video still), it’s built on a fundamentally different architecture to DALL·E and other image tools, which shows in the depth and cinematic feel of the output.
I was genuinely impressed by both Canva and Mage, not just in terms of image quality, but also how intuitive and fast they are to use. At the other end of the scale, Craiyon really didn’t deliver. Even with their so-called Pro images, the quality was noticeably lower than the others, and it struggled with a few of the prompts. The Hollywood sign example? Believe it or not, that was the best of the bunch (the only one including the correct spelling).
I deliberately chose a mix of prompts: one with a futuristic theme, one based on a typical real-world situation, and one focused on lettering, historically a weak spot for many tools, but one that’s definitely improving.
Comments