Midjourney's model updates significantly affect how SREF codes behave. Codes that produced stunning results in V6 may look different — sometimes better, sometimes unexpected — in V7. Understanding these differences helps you get consistent results regardless of which model version you're using.
What Changed Between V6 and V7 for SREF Codes
1. Style Interpretation Has Changed
V7 has a fundamentally different understanding of visual style. It's more literal and detailed in its interpretation of prompts, which means SREF codes are sometimes applied more subtly — or more aggressively — than in V6, depending on the specific code.
2. The --sv Parameter Is Now More Relevant
V7 introduced expanded support for --sv (style version), which controls how the style reference is applied. In V7, there are six style versions (sv1 through sv6), each creating meaningfully different outputs from the same sref code. V6 had limited sv options. In V7, always specify --sv explicitly to ensure predictable results.
3. Photorealism Is Stronger in V7
V7 is significantly more photorealistic by default. SREF codes that produced painterly or illustrated outputs in V6 may produce more photographic results in V7. Codes intended for illustration work may need to be combined with text prompts specifying the medium.
For V7, always add --sv 1 for the most reliable sref application, or --sv 4 for professional/commercial photography aesthetics. Don't leave --sv unspecified in V7 — the default changes between versions.
Which SREF Codes Still Work in Both V6 and V7?
Codes with strong, distinct aesthetic signatures tend to translate across versions reliably. Subtler codes that rely on specific V6 rendering characteristics may produce inconsistent results in V7. The safest approach: test your favourite V6 codes in V7 with --sv 1 first, then explore other sv values.
How to Tell If a Code Works Well in V7
Generate the same prompt three times with the same sref code in V7. If the aesthetic is consistent across all three outputs, the code translates well. If the outputs vary wildly, the code is less stable in V7 and you may need to look for an alternative with a similar aesthetic.
Using V6 Codes as a Starting Point
The SREF codes in the MidjourneyCodes library have been tested and verified — and while the samples were generated to show the aesthetic character of each code, the underlying codes work across V6 and V7. When you find a code you love, test it in your current model version and adjust --sv until you get the look you want.
Meta AI Version
Meta AI currently runs an older version of the Midjourney model — closer to V6 than V7. If you're using codes on Meta AI, expect results closer to V6 output. This is actually an advantage for codes that performed better in V6.