Playform is an AI art studio for visual thinkers. Our intuitive, code-free platform enables artists and creative professionals to harness the latest artificial intelligence technology for visual experimentation and producing new imagery.
Playform was built by a team of artists and computer scientists with a shared interest in visual culture, fine art, and emerging technology. Artists and creative professionals all over the world use our AI studio, including professors at top art schools and gallery-represented artists. Playform also runs a residency program in which we give artists resources to create and opportunities to exhibit work made with Playform. As AI continues to inform many material and aesthetic aspects of our experience, we believe it is crucial to open up access to AI-assisted imagery creation to more people.
Create an account on create.playform.io.
Browse the Explore tab so see other Playform creatives' work.
On the Collections tab, browse public image collections users have shared by clicking Public. You can create projects immediately from any collection.
On the Process tab, find one that interests you, then click Create Project to start experimenting.
Join our community on slack to ask questions and give feedback.
Artificial intelligence (AI), machine learning (ML), and deep learning—these are all terms that can be used to describe the category of technologies that power Playform. These technologies are used widely in consumer software (like Netflix’s recommendation engine), finance (like credit card fraud detection), medicine, agriculture and beyond.
A machine learning project generally involves training data that is used to teach (or train) a statistical model. The output of any project is the fully trained model. In almost all cases, providing more training data makes the resulting models more accurate. Additionally, training a model takes a lot of computation power and time, oftentimes hours or even days of running a top-of-the-line graphics processing unit (GPU).
Some trained models are used to generate data that’s similar to the training data: for example, you can teach an ML model to generate Shakespearean sonnets by training it on a ton of real sonnets). Other types of models allow you to feed it test data, and it’ll generate an output: for example, an email spam model will process each new email you receive and assign it a spam likelihood score between 0% and 100%.
Playform utilizes a particular class of ML systems called Generative Adversarial Networks (GANs), in which two artificial neural networks play a turn-based game with each other that teaches them the statistical attributes of the training data. In the case of Playform, you can think of it as a cat-and-mouse game between two digital “brains” who play the role of a forgery artist and a detective. The detective is shown the training data: a collection of “real” images that the forger is trying to imitate. Each iteration of the game, the forgery artist gets better at mimicking those images, and the detective gets better at detecting forgeries, which in turn makes the forger more skilled. After many iterations (tens of thousands or even hundreds of thousands), the forgery artist is able to make convincing imitations of the training data.
Providing more training data will allow your results to look more similar to the training data.
AI model generally don't have access to any outside information, only the training data that you input for that specific project.
Results are generated iteratively over time, and earlier results are not as well-trained as later ones.
Right now, it takes a decent amount of time and money to train a model. :-)