Hi everyone!👋🏻 If you follow me on Twitter, you saw me playing with MidJourney.. a LOT! Well, now I’ll share everything that I learned with you. Plus some other interesting things…Let’s goo 👇🏻

A journey into MidJourney 🚂

MidJourney, an independent research lab, released the fourth version of its AI algorithm.

It seems like it works impeccably well for faces (especially close-up portraits), photorealism, Pixar styles, iOS logos, and illustrations.

It also does well in terms of language comprehension, so you don't need to over-engineer the prompts to get good results.

Hands and text are still a no-go 🙅🏼‍♀️

Unfortunately, they don’t have open API, and you only get 25 free uses! If you want to generate more images, you can subscribe to the paid plans which are $10 for ~200 images, or $30 for unlimited images.

Text on AI-generated images 🏞

NVIDIA released a new text-to-media model called eDiffi, that generates better text-to-image synthesis than other models, meaning that your photo will be more accurate.

It seems that the trend for all new models is to minimize the amount of work we have to do with prompt engineering.

But that’s not even the best part.

What is unique about this model is that we’ll be able to see the correct text instead of ᏇᏋ’ᏝᏝ ᎦᎥᏁᏗᏝᏝᎩ on our results!

What else is going on

• OpenAI pack for Coda users - use Dalle-2 and GPT3 to automate busy work directly in Coda.

• Now you can productionize Large Language models using Spellbook - you just import your data, prompt or fine-tune and you can even connect it with your apps like Google Sheets via their API

• Dreambooth model that mimics the style of MidJourney, just run this Colab notebook and include “mdjrny-v4 style“ in your prompts.

• Use PromptDeck if you want to manage your prompts! I’ll need this for sure

• Check these amazing Spellbooks that Emmanuel created on top of other game assets he is creating lately.Subscribe now

This image is not real

Thank you for reading! ❤️