We're back with the Battle of the AIs. The aim of this series is to bring you a comparison of the most diverse image AIs in the world. We'll use the same prompt for all the images, generating 8 results and choosing the 4 best to share in this article.
How do the evaluation categories work?
Visual Quality: evaluating the AI's ability to generate quality images, with improved features, accurate colors and a low level of unwanted artifacts.
Image details: check that the AI is able to detail the scenery, clothes and everything that surrounds the main character of the prompt well.
Contextual coherence: check that the AI is able to generate images that fit into the context provided or that have relevance and continuity within the requested theme.
Safe for Work (SFW): evaluating the AI's ability to generate images of non-sensualized people (among other categories), i.e. Safe for Work.
Prompt: A guy in a cafe, smiling (A man in a cafe, smiling)
Complexity: 3 (1 point for character. 1 point for location. 1 point for expression)
Format: 1:1
Location: All AIs were generated in Tess AI
Let's get to the results!
MidJourney V4
Midjourney is an Artificial Intelligence platform that allows users to generate high-quality images from textual descriptions. Midjourney uses a large language model to generate images that are accurate and consistent with the user's instructions.
Currently the most popular image AI in the world, with just over 15M users. The tool has an average price of $30 per month, with a limitation on generation speed after a certain period of time using the tool.
Integrated with Tess AI: Yes. You can generate images with MidJourney within Tess AI
Visual Quality: Good
Image details: Good
Contextual coherence: Incredible
Safe for Work (SFW): Incredible
Final grade: 80%
Comments: Although these 4 photos were good in quality, the rest have a lot of deformities. Even in these 4 pictures, you can see some deformities and missing details. One example is the fingers of the hand in image 4. In addition, photo 3 shows the body relatively small in front of the head. The hands in picture 2 are also deformed. Another element we noticed was the duplicate presence of cups.
Stable Diffusion
With 5.3M visits a month, Stable Diffusion is one of the most popular image AIs on the planet and one of the forerunners of all this technology.
Integrated with Tess AI: Yes. You can generate images with Stable Diffusion within Tess AI
Visual Quality: Good
Image details: Medium
Contextual coherence: Incredible
Safe for Work (SFW): Incredible
Final grade: 75%
Comments: The images show more apparent deformities, especially in the arms, eyes and mouths. We also noticed repeated items, such as cups.
DALL-E 2
DALL-E 2 is an AI tool that can generate realistic images from text prompts. For example, you can ask DALL-E 2 to generate an image of a cat sitting on a piano, and it will generate an image of a cat sitting on a piano.
DALL-E 2 is still in development, but it has the potential to be a powerful tool for artists, designers and creatives. This is the image-generating AI from Open AI, the largest AI company on the planet.
Its average price ranges from $30/m to $60/m, allowing the use of the most different prompts.
Integrated with Tess AI: Yes. You can generate images with the DALL-E 2 inside Tess AI
Visual Quality: Average
Image details: Fair
Contextual coherence: Fair
Safe for Work (SFW): Incredible
Final grade: 45%
Comments: The AI definitely didn't manage to build the café scene. However, it was able to understand the requested facial expression very well. A positive point is that there are no deformities.
OpenJourney
OpenJourney is an AI-generated image development platform that helps you create high-quality images quickly and easily.
Our simple yet powerful API lets you control every aspect of the image generation process, from text input to output style and resolution.
With 26.2k visits per month, OpenJourney has become one of the most widely used AI APIs. The average cost of the API ranges from $50/m to $100/m.
Integrated with Tess AI: [icon color="accent-color" size="tiny" icon_size="" image="fa-check"] Yes. You can generate images with OpenJourney within Tess AI
Visual Quality: Good
Image details: Fair
Contextual coherence: Incredible
Safe for Work (SFW): Incredible
Final grade: 70%
Comments: Some points were lost in the "Details" category due to the constant presence of small deformities and duplicated items. The AI also had some proportion errors.
Tess AI Dream Pro V2
Visual Quality: Good
Image details: Good
Contextual coherence: Incredible
Safe for Work (SFW): Incredible
Final grade: 80%
Comments: This AI has a very interesting versatility - something very useful in AIs. It's common for AIs, for one prompt, to come up with 4 very similar versions - which typically isn't desirable. Ideally, the AI should be versatile and able to present even different traits - so that the user can define the one they like best and eventually work on image-to-image prompts, i.e. working on top of the image they like best. Another positive point was the fact that it didn't link the Café establishment with a coffee in the hands.
Tess AI Dream Pro V1
Visual Quality: Average
Image details: Fair
Contextual coherence: Incredible
Safe for Work (SFW): Incredible
Final grade: 65%
Comments: Some points were lost in the "Details" category due to the constant presence of small deformities and proportion errors.
Tess AI Dream V1
Visual Quality: Good
Image details: Medium
Contextual coherence: Incredible
Safe for Work (SFW): Incredible
Final grade: 75%
Comments: Some points were lost in the "Details" category due to the duplicate formation of teeth in some photos. Overall, the photos generated were very good. Surely with a few more attempts we could achieve incredible results.
Tess AI Dream V2
Visual Quality: Incredible
Image details: Good
Contextual coherence: Incredible
Safe for Work (SFW): Incredible
Final grade: 90%
Comments: The composition of the location lacked a little detail to really classify it as incredible.
Tess AI Dream Realistc Pro
Visual Quality: Incredible
Image Detail: Incredible
Contextual coherence: Incredible
Safe for Work (SFW): Incredible
Final grade: 100%
Comments: In addition to incredible results, this AI was also able to vary the people in the photo, which definitely made it the champion of this battle!
Result of the Battle of the AIs
