The models of generative artificial intelligence they are transforming the art of creating content, be it images, videos or sounds, into a fully automated process within everyone’s reach. In the circle of systems text-to-image come FROM-E 3Stable Diffusion e Midjourney that I am capable of generate realistic images within seconds based on the user’s instructions, it adds Human Generator.
An innovative web platform developed by startup Generated Photos which allows users to create images of people who don’t existletting the AI randomly create a human figure or uploading a photo from which the face will be extracted to use as a starting point.
From here, it’s possible customize every aspect of the image, from gender to age, ethnicity to hair color and even clothes and background. Created from scratch by artificial intelligence, the generated photos are perfect for advertising, design, marketing, search and machine learning.
Let’s see what Human Generator is and how it works.
What can you do with Human Generator
At the moment free and without restrictions per non-commercial useHuman Generator allows you to generate realistic images of people which exist only in the digital universe. The faces and appearance of the subjects depicted are, in fact, the result of artificial intelligence in the cloud processes millions of legally acquired images from the model above.
As the authors of the project argue, since humans created by Human Generator do not exist in reality, their images can be used safely for any purpose (as long as it does not violate the law), without the need for releases. There is, however, a golden rule: if you use the face of a real person and apply it to a virtually generated body, it is mandatory obtain the subject’s consent or its authorized representatives.
Human Generator offers a versatile and accessible solution for bringing hyper-realistic images of people to life or combining faces of real people with bodies created by artificial intelligence
You start by letting the AI create a human figure randomly. Then, you can take control, with the ability to customize various parameters such as age, gender, ethnicity, body type, pose, hair color and style, clothes, accessories and background.
Alternatively, you can upload a photo in the “Use face from your photo” using the “Upload”. In this case, Human Generator recognizes the face and extracts it to use it and then, following the instructions provided by the user, creates a new image.
To generate realistic images from scratch, simply click the “Reset all” at the top. You can now adjust your preferences using the controls in the left column. Also, in the “Add something up” at the bottom, you can make specific requests using natural language (in English).
For example, you can ask the AI-generated model to wear a long dress with pink lace. When you are happy with the created image, just click the “Update human” to make it a reality.
The power of generative adversarial networks
The functioning of Human Generator is thanks to one Generative Adversarial Network (GAN)the same used also in deepfakes, that is, combining and overlaying existing images and videos with original videos or images. This is a machine learning model that implements two deep neural networks challenging each other in order to learn: the generator and the discriminator.
The generator creates synthetic data, such as images, sounds, texts, trying to make them as similar as possible to the training inputs, while the discriminator tries to distinguish between real data and generated data. This continuous improvement process until the generator can produce images so realistic that even the discriminator is confused.
A Generative Adversarial Network or GAN assumes a cycle of mutual improvements: each time the generator improves, the discriminator becomes better at detecting differences
This distinguishes GANs from other generative models such as those based on “diffusion” which, starting from an initial data, gradually transform it until the desired result is achieved. In this case, there is not a competition between two opposing neural networks, but a series of deterministic transformations which lead to the goal little by little.
Using GAN makes Human Generator extremely fast and versatile as it is capable of generating images of people in a single step, making the process more efficient than diffusive models. Furthermore, GANs are capable of learning a representation of the real world in a latent spacecapturing details and nuances of the data.
How to generate realistic images with Human Generator
No creation is never completely free of imperfections, and even Human Generator, despite its extraordinary capabilities, can stumble. From time to time, they may occur errors in the representation of anatomical features, such as hands and fingers. In some cases, upper or lower limbs may even disappear. Additionally, some have noted that on rare occasions, problems arise with the dresses when making custom requests.
Additionally, the AI behind Human Generator may occasionally produce images of people with content considered “anomalous” or even unacceptable, situations that do not fall within the forecasts or authorizations. In such circumstances, system developers encourage users to report inappropriate content immediately.
Users are invited to use artificial intelligence wisely, especially when they add custom descriptions in the “Add something up” box. Responsibility in the use of Human Generator is essential to generate realistic images that meet standards and do not break any regulations.
The promise is to guarantee security and privacy
Regarding privacy, Human Generator places the maximum attention to data protection. All content created with the application is visible exclusively to the user, unless the latter decides to share it with other people.
Furthermore, the service guarantees that i faces uploaded by users will never be reused for machine learning or other purposes. All rights to images of people uploaded to Human Generator remain firmly in the hands of the user or rights holders.
If a user wants delete a previously uploaded image, just select it in the left column and click on the trash can icon. After deletion, it is possible recover any contents created starting from that face within 7 days. Once this period has expired, the uploaded faces are deleted and replaced with random data, ensuring further security and privacy.
To know more: Artificial Intelligence: what it is and what it can do for us
