Midjourney v6 Prompt Changes: What I Rewrote and Why
Midjourney v6 changed how it interprets prompts. Here is what I rewrote in my prompt library and the patterns that still work. Midjourney v6 was not an incremental update. It changed how the model interprets prompts, which meant every prompt I had refined for v5 needed reevaluation. Some prompts that produced great results in v5 produced flat or broken results in v6, and a few patterns that v5 ignored suddenly became powerful. After reworking my entire prompt library, I have a clear picture of what changed, what to rewrite, and what still works. This guide covers the changes that affected my workflow the most. Natural Language Over Comma Lists In v5, I wrote prompts as comma-separated keyword lists: portrait, woman, soft light, film grain, 85mm. The model treated each keyword as an independent tag and blended them. In v6, the model reads prompts more like natural language, and comma lists produce weaker results than full sentences. A v6 prompt that reads as a coherent description outperforms a keyword list of the same elements. // v5 style (weaker in v6) portrait, woman, soft light, film grain, 85mm, shallow depth of field // v6 style (stronger) A portrait of a woman lit by soft north-facing window light, shot on 85mm film stock with shallow depth of field and visible film grain The sentence structure lets the model understand relationships between elements.