Introducing Playground v3
Our new model's focus was to be the best at prompt understanding and control—going beyond aesthetics which has saturated as a benchmark. It outperforms all the most popular image foundation models in its class.
Prompt understanding
Superhuman graphic design abilities
In evaluations across popular graphic design categories, users consistently chose PGv3's designs over the similar human-made ones. A sample from those evals is below.
Playground v3 win rate
Creating an AI graphic designer required new capabilities
Our research focused on improving control over visual communication, typography, color precision, and layout.
PGv3 excels at generating accurate text in context, making it ideal for everyday design tasks. With a text-synthesis score of 82%, PGv3 outperforms all other SOTA image models on text generation.
Edit complex designs naturally
A great graphic designer blends cultural knowledge, aesthetics, and design principles.
PGv3 shines in all these areas, thanks to its LLM-integrated structure. It understands and follows detailed composition, layout, and style directions, while also grasping cultural references like holidays, memes, celebrities, sports teams, and more.
Argus is a new vision language model that can describe images better than a human could (or would want to).
VLMs are a key component to improving our image model so that people to control every element of a design.
precisely