Midjourney releases V6 β overhauls in-image text rendering and prompt interpretation
Midjourney has released an alpha version of its new model βV6,β trained from scratch over nine months. The headline feature is the ability to render legible text within images, and the way prompts are interpreted has also been overhauled significantly compared to V5 and earlier.
Details
- In-image text: Short text placed in quotation marks within a prompt can now be rendered as legible text inside the generated image
- Overhauled prompt interpretation: Stock modifiers like βaward winning,β βphotorealistic,β and β4kβ are no longer necessary, and the model now responds better to more specific, natural-language instructions
- Long prompt support: Longer, more detailed prompts are now reflected more accurately
- How to use it: Made available via the
/settingscommand in Discord, or by appending--v 6to the end of a prompt
What happened next
V6 became the default model on February 14, 2024, and remained the standard model until being succeeded by V6.1 on July 30 that same year. By overhauling how prompts should be structured, it marked a turning point significant enough that the user community had to relearn how to write effective prompts.