Paracosmic Explorers' SocietyLots of exciting things happening in the proprietary image generation space nowadays. First, I've been given preview access to #Reve, a startup company with an image model that has a very high quality and prompt adherence. Its strongest suit, however, is accurate and beautiful text and typography. Just look at this:
The really big thing this week, however, is multimodal image generation. It's hard to overstate what big leap this is in terms of prompt adherence, quality, and just overall versatility. Gemini made it available first, although it's kind of hidden in an inaccessible user interface. But it was OpenAI that brought it to the mass market, with native image generation in #ChatGPT.