Update: Even John Oliver has now published a video about this: https://www.youtube.com/watch?v=3YNku5FKWjw
Tl;dr: AI can generate great images from a text prompt. Try it yourself!
Artificial intelligence (AI) is already used in so many different areas, such as screening applicants’ CVs, fraud detection in financial transactions, or giving recommendations for your next streaming series. But it is also good at producing content, be it text via GPT-3 and it’s relatives or even images based on a text prompt. Google’s ImageGen is a good example, and it has deliberately not been made puclicly available because of the dangers of abuse and misinformation. Other offers have partially closed or open betas available.
One of my favourites is MidJourney. They recently went into public beta, so everybody can sign up and toy around with it in order to generate images from text. In order to use it, you need to sign up for the beta, then get an invite to a Discord server where a bot will gladly accept your prompts and generate images for you. For results you can then choose to make variants or higher-resultion versions.
If you’re interested, you can find MidJourney here:
- WebSite: https://www.midjourney.com/
- Direct link to Discord: https://discord.com/invite/midjourney
- Handbook with usage instructions: https://midjourney.gitbook.io/docs
Important hint: With the ’letter’ emoji on a result, you can instruct the chat bot to send it to you via direct message. There’s a lot of traffic in those channels, so it may be easy to lose track of where your own results were.
Who wants to try image generation without signing up for Discord can use the publicly available alternative Crayion:
And of course, there’s DALL-E 2 itself, but you still need an invite:
Which great AI tools to you know and use? 🐉🔥