Everyone has had a dream, an inspiration, or a mental picture that they want to share and cannot wait to tell others about but, for most people, expressing it through traditional artistic mediums such as painting can be challenging. How amazing would it be if we could replicate the images in our mind using only words and bring them to life through artwork. Well, we’re developing an application called Vieutopia that allows you to create unique and personalised high-resolution artwork using only your phone.
The autoencoders and diffusion probabilistic models are combined to create our backbone. Similar to humans, artificial intelligence models require sufficient and appropriate data to learn. Over 6 billion filtered image-text pairs, containing LAION-5B1 and our self-constructed dataset, are used for testing. Following months of intense testing, our program can generate unique images that offer a realistic look.
Our model is more suitable for new users and generates more artistic artwork with higher resolution than other models like DALL-E 2 and Stable Diffusion. Even users who are unfamiliar with our application can create amazing works of art! A deeper understanding of our models can generate truly incredible images with more elaborate prompts. There are tons of amazing text-to-image applications available, however, they are simply focused on generating the images which fit the prompts. Those images cannot be called art, and people can simply recognise they are generated by artificial intelligence, thus people cannot feel the beauty and connection to art. However, our application can generate indistinguishable and various styled artworks from different time periods. The generated works have vivid strokes and appropriate colour tones to offer a truly realistic and lifelike look.
Our intention is to make this new technology accessible to everyone, not to replace artists. The goal is to allow more people to appreciate the beauty of art and the fulfilment of bringing their own creativity to life. We can’t wait to see all the incredible personalised images everyone will create with our application. We hope you have fun unleashing your creativity!
Henry Qu is an Algorithm Engineer for Vieunite. His background is in Artificial Intelligence and Machine Learning.