Midjourney 5.1
4th May 2023
Midjourney released version 5.1 of their image generation model on Tuesday. Here’s their announcement on Twitter—if you have a Discord account there’s a more detailed Discord announcement here.
They claim that “V5.1 is more opinionated (like V4) and is MUCH easier to use with short prompts”—in comparison to v5.
Last night (9:30pm PST on Wednesday May 3rd) they switched 5.1 to be the default—previously you had to add --v 5.1
to a prompt in order to use it.
To compare the v5 and v5.1 models, I ran the prompt pelicans having a tea party through them both.
Midjourney v5
v5 is the version of Midjourney that came out on March 15th, and really felt like a turning point in that it was the first to reliably produce photorealistic images. If you’ve seen the flurry of memes of the Pope in a Balenciaga puffy jacket, you’ve seen Midjourney 5.
Midjourney v5.1
I find the difference between the two so interesting. The v5 one went for photo-realism—the pelicans are in a natural setting, and while they are standing near a tea service none of them are really interacting with it beyond looking at it.
For 5.1, the model seems to have made very different choices. These pelicans are in a formal setting—a tea room, albeit in some with an oil painting of the ocean behind them. The style is more illustrative than photographic, and definitely more whimsical. They’re interacting with the tea—which means the model as added creepy little hands in three cases and in one case given them pink claws, albeit in addition to their existing wings.
I think 5.1 does a better job with this admittedly vague and silly prompt.
I use Midjourney pretty regularly now, exclusively for entertainment. It’s a lot of fun.
More recent articles
- Weeknotes: more datasette-secrets, plus a mystery video project - 7th May 2024
- Weeknotes: Llama 3, AI for Data Journalism, llm-evals and datasette-secrets - 23rd April 2024
- Options for accessing Llama 3 from the terminal using LLM - 22nd April 2024
- AI for Data Journalism: demonstrating what we can do with this stuff right now - 17th April 2024
- Three major LLM releases in 24 hours (plus weeknotes) - 10th April 2024
- Building files-to-prompt entirely using Claude 3 Opus - 8th April 2024
- Running OCR against PDFs and images directly in your browser - 30th March 2024
- llm cmd undo last git commit - a new plugin for LLM - 26th March 2024
- Building and testing C extensions for SQLite with ChatGPT Code Interpreter - 23rd March 2024
- Claude and ChatGPT for ad-hoc sidequests - 22nd March 2024