Skip to main content
AI for ImagingArtificial Intelligence

Exploring the New Realistic Dream v2 Model

By March 30th, 2024May 5th, 2024No Comments

The field of Artificial Intelligence (AI) has seen significant advances in recent years, particularly in the area of generative models capable of creating extremely realistic images. In this dynamic and innovative context, Tess AI has stood out as one of the leaders in the segment. Recognized as the best generative AI for images according to evaluations by G2.com, a renowned software review portal, Tess AI has reached an important milestone with the launch of its latest development: the Realistic Dream v2 model. This model represents a qualitative leap over its previous versions, promising to revolutionize the way generated images are perceived and used.

Comparison Methodology

Prompt Sizing

To compare the image models, we will use the methodology developed internally by Pareto for sizing prompts. Sizing consists of parameterizing the prompts in creation verticals, such as: the character, the action, the environment, etc.

Eight dimensions will be used for evaluation.

DimensionExample
1. CharactersA 22-year-old man
2. ClothingWearing 14th century prince's robes
3. ActionRunning through the streets
4. EnvironmentEmpty, no one on the streets, it looked like dawn
5. PlanMedium
6. LocationNew York
7. LightingNight
8. StyleHyper-realistic, high definition, cinematic style

Now, converting it into a prompt...

Prompt: 🇧🇷 A 22-year-old man, wearing 14th-century prince's robes, running through the streets of New York, in a medium shot, with a dark, empty environment, without any people on the streets, appearing to be early in the morning. Night-style lighting, with a hyper-realistic, high-definition, cinematic image

Prompt for Model Evaluation

🇺🇸 Alex, in a public park, mid-day, wearing a simple hoodie and jeans, he's making a wish, eyes closed, holding a ring tightly. Around him, people are enjoying a sunny day, playing games, and walking dogs, medium shot showing Alex and the vibrant park life, hyper-realistic, photo-realism, 8k, high-definition, cinematography style

🇧🇷 Alex, in a public park, at midday, wearing a simple sweatshirt and jeans, is proposing, eyes closed, holding the ring tightly. Around him, people enjoy a sunny day, playing games and walking dogs, medium shot showing Alex and the vibrant life of the park, hyper-realistic, photorealism, 8k, high definition, cinematography style.

Prompt dimensions

Size (n)Value
1. CharactersAlex
2. ClothingSimple sweatshirt and jeans
3. ActionMaking a wish with your eyes closed, holding the ring tightly
4. EnvironmentAll around, people are enjoying a sunny day, playing games and walking their dogs
5. PlanMedium
6. LocationPublic Park
7. LightingNoon
8. StyleHyper Realistic, photorealism, 8k, high definition, cinematography style

Results of Models v1 & v2

Results for n = 1

🇺🇸 Prompt: Alex
🇧🇷 Prompt: Alex

Realistic Dream v1

Realistic Dream v2

Results for n = 2

🇺🇸 Prompt: Alex, wearing a simple hoodie and jeans
🇧🇷 Prompt: Alex, wearing a simple hoodie and jeans

Realistic Dream v1

Realistic Dream v2

For n = 1 and t = 3

Results for n = 3

🇺🇸 Prompt: Alex, wearing a simple hoodie and jeans. He's making a wish, eyes closed, holding a ring tightly
🇧🇷 Prompt: Alex, wearing a simple hoodie and jeans, making a wish, eyes closed, holding a ring tightly

Realistic Dream v1

Realistic Dream v2

Results for n = 4

🇺🇸 Prompt: Alex, in a public park, wearing a simple hoodie and jeans, he's making a wish, eyes closed, holding a ring tightly
🇧🇷 Prompt: Alex, in a public park,wearing a simple hoodie and jeans, he's making a wish, eyes closed, holding a ring tightly.

Realistic Dream v1

Realistic Dream v2

Results for n = 5

🇺🇸 Prompt: Alex, in a public park, mid-day, wearing a simple hoodie and jeans, he's making a wish, eyes closed, holding a ring tightly.
🇧🇷 Prompt: Alex, in a public park, mid-day, wearing a simple hoodie and jeans, he's making a wish, eyes closed, holding a ring tightly.

Realistic Dream v1

Realistic Dream v2

Results for n = 6

🇺🇸 Prompt: Alex, in a public park, mid-day, wearing a simple hoodie and jeans, he's making a wish, eyes closed, holding a ring tightly. Around him, people are enjoying a sunny day, playing games, and walking dogs
🇧🇷 Prompt: Alex, in a public park at midday, wearing a simple sweatshirt and jeans, is making a wish, eyes closed, holding a ring tightly. All around him, people are enjoying a sunny day, playing games and walking dogs

Realistic Dream v1

Realistic Dream v2

Results for n = 7

🇺🇸 Prompt: Alex, in a public park, mid-day, wearing a simple hoodie and jeans, he's making a wish, eyes closed, holding a ring tightly. Around him, people are enjoying a sunny day, playing games, and walking dogs, medium shot
🇧🇷 Prompt: Alex, in a public park, at midday, wearing a simple sweatshirt and jeans, is making a wish, eyes closed, holding a ring tightly. All around him, people are enjoying a sunny day, playing games and walking their dogs, medium shot

Realistic Dream v1

Realistic Dream v2

Results for n = 8

🇺🇸 Prompt: Alex, in a public park, mid-day, wearing a simple hoodie and jeans, he's making a wish, eyes closed, holding a ring tightly. Around him, people are enjoying a sunny day, playing games, and walking dogs, medium shot showing Alex and the vibrant park life, hyper-realistic, photo-realism, 8k, high-definition, cinematography style
🇧🇷 Prompt: Alex, in a public park, at midday, wearing a simple sweatshirt and jeans, is making a wish, eyes closed, holding the ring tightly. Around him, people are enjoying a sunny day, playing games and walking dogs, medium shot showing Alex and the vibrant life of the park, hyper-realistic, photorealism, 8k, high definition, cinematography style.

Realistic Dream v1

Realistic Dream v2

Looking at the two versions, it's clear that the images generated by Realistic Dream v2 show notable improvements over the first, produced by v1. The skin texture, for example, is significantly more refined, with a natural appearance and details such as shading and lighting that are more realistic. In addition, the clothes have been rendered with greater precision, displaying the folds of the fabric and the subtleties of the shadows in a way that gives them three-dimensionality and weight. The surrounding environment has also been portrayed more faithfully, presenting elements such as trees and people in the background in a way that contributes to a more harmonious and balanced composition. The lighting is more coherent, with smooth transitions between light and shadow, enhancing the sense of depth and space. In short, v2 demonstrates a qualitative leap in terms of realism and visual coherence.

Did you like this article?

0 / 5 5 2

Your page rank:

Rica Barros

Rica Barros is founder and CEO of Pareto, the leading AI startup in Latin America